2040 structural model

One need · one readable arc

Buying stops being one kind of event.

The shift is not an agent clicking today’s checkout faster. Intent becomes portable authority, businesses become callable, specialized agents coordinate the market and purchases become policies that can reach the physical world.

The human does not disappear. Attention moves from assembling transactions to setting boundaries, reviewing exceptions and revoking authority. Read the purchasing story first; expand the economic, protocol and ecology layers afterward.

Build the path

Choose what becomes true—then live through the result.

Authority, market formation, assurance and execution are independent forks. The selected care coordination need stays constant so the changing purchasing experience remains visible.

Compare three institutional benchmarks—not the full possibility space

Coordinated openportable authority + compatible rails Fragmented openadapters + partial portability Walled gardensintegrated platforms + captive trust

Published path · structural model

Coordinated open rails

Rails can bridge buyer instruments. A wallet or card token can resolve the same 402 through an interoperable router; chat runtimes rarely block a third-party rail and platform take stays near card interchange. By 2029Q2 24% of agent attempts succeed, agents carry 22% of online GMV — inside Bain's 2030 band a year early — and a third of buyers hold a wallet. The binding constraint flips from the buyer's instrument to seller reach.

2040 machine counterparties54%

continuous commerce23%

published 2040 p10–p9018%–48%512 structural draws

The whole path in four changes

One need moves from request to operating policy.

story first · model layers after

2029 · assisted commerceIntent becomes a mandate

The person gives software a bounded right to act.

purchase: approval bundle
agent-executed: $121B–$187B
machine cast: 3 role classes
human: operator

2032 · delegated commerceBusinesses become callable

Catalog, checkout, fulfillment and support become machine-addressable capabilities.

purchase: approval bundle
agent-executed: $230B–$450B
machine cast: 7 role classes
human: approver and recourse owner

2036 · agent marketsThe market gains counterparties

Specialized buyer, seller, trust and operating agents coordinate the purchase.

purchase: negotiation room
agent-executed: $398B–$935B
machine cast: 10 role classes
human: policy setter and exception handler

2040 · agent marketsThe purchase becomes a policy

Standing intent allocates money and physical capacity until a boundary is crossed.

purchase: policy console
agent-executed: $601B–$1.6T
machine cast: 11 role classes
human: policy setter and exception handler

Conditional annual ranges overlap and are not valuations. Role classes are functional activations above the disclosed threshold—not agent population or forecast headcount.

2029

assisted commerce · model-triggered authored event

Checkout compatibility exposes the seller gap.

Compatible rails let more buyer agents reach sellers, but seller-side delegation remains a narrow foothold.

2029

Illustrative Care coordination purchase · Coordinated open rails

One intent, one approval

State the goal, inspect the evidence and approve this purchase.

approval bundle · engine selected

Agent influenced$332B–$380B

Agent executed$121B–$187B

Merchant-routed after tolls + loss$107B–$168B

CCAuthorization bundleCare coordination · authenticated workspaceready for approvalCoordinated open rails2029

IntentFind three eligible care providers with the right specialty, availability and language fit; do not book without approval.

Recommended bundleEligible providers · language fit · verified availabilityCoverage, credentials and continuity evidence remain inspectable · portable mandate · independent assurance

within mandate

FallbackEarlier intake · different locationNever booked without explicit provider approval · present a portable receipt to an independent assurer, unwind the commitment and reroute

person decides

One signed commitment · one-shot settlementAuthorize purchase

terms verifiedmandate · CARE-2029-04proof attached

Illustrative outcomeThe person approves one bounded commitment and receives a receipt plus execution watch.

intentsourcenegotiatesettleassure

personsharedagent

Authoritysingle-use mandate

Market topologyfederated market

Identityportable mandate

Price formationposted bundle

Settlementone-shot settlement

Human attentionapprove this purchase

Trust surfaceportable mandate, independent assurance, provenance and a receipt that records the mandate and result.

How the market respondsAssembles a ranked offer, counterparty, settlement and assurance bundle, then waits for one explicit authorization. Sourcing uses ranked shortlist from a federated market; posted bundles set the price before one-shot settlement executes the commitment.

When it breakspresent a portable receipt to an independent assurer, unwind the commitment and reroute; the person’s attention mode is “approve this purchase.”

autonomy7%

delegation fit13%

market reach72%

exception risk28%

Range-covering interface spread

approval 100%negotiate 0%policy 0%ambient 0%

512 structural draws · sensitivity coverage, not probability

Surface sensitivityThe engine-selected surface appears in 100% of the range-covering draws above. The remaining share shows which alternate purchasing interfaces the same model selects as structural assumptions move.

Evidence labelIndicators combine structural state with disclosed, unanchored arena priors. Dialogue and interface remain illustrative, not forecast outputs.

In the physical world · illustrative physical-world experience

The agent prepares the visit; the person chooses the provider.

2029 · Coordinated open rails
federated market
Site referral 77%

What the person experiencesThe visible moment

The person sees three eligible providers with verified specialty, language fit, availability and coverage evidence.

What software does off-screenThe delegated work

The agent verifies administrative facts, resolves contradictory directories and prepares—but does not submit—the booking.

What changes in the worldThe physical handoff

After explicit approval, the clinic receives the minimum eligibility and scheduling information needed for intake.

Control returns hereProvider selection, booking and every clinical choice remain with a person.

What happens next · model-derived interpretation

The visible change is compression: product search, seller choice, service, payment and delivery arrive as one proposal. Human attention concentrates at the moment of authorization. At the 2029 handoff, 22% of demand is delegated while only 2% has a machine on both sides. The experience therefore begins as an approval bundle.

What remains contested: A federated market expands choice, but portable mandate and independent assurance must scale without hardening into a new control point.

Active machine cast · model-derivedBuyer agent · Mandate agent · Settlement agent

Coordination burden2%range 2%–3%

Human roleoperatorowns multi-agent drift

Between purchasing chapters · model-derived experience transition

The structural handoff begins as an approval bundle.

2029Q2 → 2029

Interfaceapproval bundlestructural starting state

Human attentionapprove this purchasestructural starting state

Agent mandatesingle-use mandateautonomy starts at 7%

Market boundary72% reachablereach starts at 72% · exception risk starts at 28%

Model trace · why this chapter changes

Model trigger · branch-specific event at the near-term handoff21.9% of demand is delegated, while seller-agent adoption begins at 7.8% after the 2029Q2 handoff.Authored consequence: Merchant representation—not buyer intent—becomes the binding work for the open ecosystem.

Event sensitivity · branch-fixed handoff event

Checkout compatibility exposes the seller gap.100%

512 range-covering structural draws · authored event selection coverage, not probability. Checkout compatibility exposes the seller gap. 100%

delegated demand22%

machine on both sides2%

continuous commerce2%

physical execution0%

pooled demand0%

adaptive supply0%

governance capacity76%

adversarial loss8%

2029Q2 → structural 2029model-derived transition · equations, not observed attribution

delegated demandThe structural layer carries the branch’s final agent-GMV share into its starting delegated-demand state.22%near-term agent GMV 22% → 22%

seller-agent footholdA disclosed 12% handoff factor converts rail-capable sellers into the initial seller-agent foothold.8%sellers answering a rail 65% → 8%

governance capacityPortable authority, liability clarity and branch coordination are averaged into the initial governance capacity.76%structural priors

physical execution footholdA disclosed handoff factor seeds the small share of agent demand that can already cause and verify physical execution.0%near-term agent GMV + structural prior 22% → 0%

pooled demand footholdDelegated demand and the pooled-mandate prior seed the initial ability to bargain across more than one buyer.0%agent GMV + pooled-mandate prior 22% → 0%

adaptive supply footholdA disclosed handoff factor seeds supply that can be configured from machine-readable demand rather than selected from a fixed catalog.0%seller-agent foothold

adversarial lossThe failed-attempt share and adversarial-pressure prior seed loss pressure; this is a structural mapping, not observed loss.8%failed attempts + pressure prior 76% → 8%

Arena lens: For care coordination, repeatability 55%, reversibility 25%, consequence 90% and preference-articulation load 75% produce 13% modeled delegation fit. These are unanchored scenario priors, not measurements.

share of demand delegated starts at 21.9%; seller-agent adoption starts at 7.8%; share with machine counterparties starts at 1.7%; share negotiated starts at 1.8%; share run continuously starts at 2.2%.

No new actor class enters yet; adoption and institutional state drive this transition.

modeled from unanchored structural inputs · 17 published research hooks · ensemble ranges are modeled sensitivity, not measured probability

2032

delegated commerce · model-triggered authored event

Mandates become portable credentials.

Authority, receipts and reputation begin crossing runtimes without forcing the buyer to create a new platform account.

2032

Illustrative Care coordination purchase · Coordinated open rails

One intent, one approval

State the goal, inspect the evidence and approve this purchase.

approval bundle · engine selected

Agent influenced$555B–$862B

Agent executed$230B–$450B

Merchant-routed after tolls + loss$209B–$415B

CCAuthorization bundleCare coordination · authenticated workspaceready for approvalCoordinated open rails2032

IntentNegotiate intake timing and verified coverage while preserving provider choice and continuity constraints.

Recommended bundleEligible providers · language fit · verified availabilityCoverage, credentials and continuity evidence remain inspectable · portable mandate · independent assurance

within mandate

FallbackEarlier intake · different locationNever booked without explicit provider approval · present a portable receipt to an independent assurer, unwind the commitment and reroute

person decides

One signed commitment · conditional atomic settlementAuthorize purchase

terms verifiedmandate · CARE-2032-04proof attached

Illustrative outcomeThe person approves one bounded commitment and receives a receipt plus execution watch.

intentsourcenegotiatesettleassure

personsharedagent

Authoritysingle-use mandate

Market topologyfederated market

Identityportable mandate

Price formationnegotiated bundle

Settlementconditional atomic settlement

Human attentionapprove this purchase

Trust surfaceportable mandate, independent assurance, provenance and a receipt that records the mandate and result.

How the market respondsAssembles a ranked offer, counterparty, settlement and assurance bundle, then waits for one explicit authorization. Sourcing uses ranked shortlist from a federated market; buyer and seller agents negotiate the bundle before conditional atomic settlement executes the commitment.

When it breakspresent a portable receipt to an independent assurer, unwind the commitment and reroute; the person’s attention mode is “approve this purchase.”

autonomy15%

delegation fit13%

market reach72%

exception risk26%

Range-covering interface spread

approval 99%negotiate 1%policy 0%ambient 0%

512 structural draws · sensitivity coverage, not probability

Surface sensitivityThe engine-selected surface appears in 99% of the range-covering draws above. The remaining share shows which alternate purchasing interfaces the same model selects as structural assumptions move.

Evidence labelIndicators combine structural state with disclosed, unanchored arena priors. Dialogue and interface remain illustrative, not forecast outputs.

In the physical world · illustrative physical-world experience

Administrative terms negotiate; care decisions do not.

2032 · Coordinated open rails
federated market
Site referral 51%

What the person experiencesThe visible moment

The person sees an earlier intake at a new location beside a later appointment that preserves continuity and language fit.

What software does off-screenThe delegated work

Provider, benefits and transport agents negotiate eligibility, intake timing and logistics without exchanging clinical discretion.

What changes in the worldThe physical handoff

The chosen clinic, ride and benefits authorization are coordinated only after the person selects the provider path.

Control returns hereThe mandate excludes diagnosis, treatment and autonomous provider substitution.

What happens next · model-derived interpretation

The visible change is compression: product search, seller choice, service, payment and delivery arrive as one proposal. Human attention concentrates at the moment of authorization. Since 2029, machine-counterparty share rises by 10.6 percentage points and continuous commerce rises by 1.3 percentage points. The experience remains an approval bundle, but its mandate and market scope deepen.

What remains contested: A federated market expands choice, but portable mandate and independent assurance must scale without hardening into a new control point.

Active machine cast · model-derivedBuyer agent · Seller agent · Mandate agent · Settlement agent · Routing agent · Procurement agent · Market operator

Coordination burden3%range 2%–5%

Human roleapprover and recourse ownerowns multi-agent drift

Between purchasing chapters · model-derived experience transition

approval bundle remains visible while its mandate deepens.

2029 → 2032

Interfaceapproval bundleunchanged from approval bundle

Human attentionapprove this purchaseunchanged from approve this purchase

Agent mandatesingle-use mandateautonomy +8.1 pp to 15%

Market boundary72% reachablereach +0.1 pp to 72% · exception risk -2.3 pp to 26%

Model trace · why this chapter changes

Model trigger · identity portability ≥ 65.0%Identity portability is 78.0% and machine counterparties reach 12.3%.Authored consequence: Assurance networks now decide which permissions and histories can safely travel with an agent.

Event sensitivity · selected 31% · leading alternate 38%

Open rails outrun portable authority.31%

Assurance bridges begin translating authority.38%

Mandates become portable credentials.31%

512 range-covering structural draws · authored event selection coverage, not probability. Open rails outrun portable authority. 31% · Assurance bridges begin translating authority. 38% · Mandates become portable credentials. 31%

delegated demand42%

machine on both sides12%

continuous commerce3%

physical execution7%

pooled demand14%

adaptive supply3%

governance capacity76%

adversarial loss5%

2029 → 2032model-derived transition · equations, not observed attribution

delegated demandCommercial capability closes part of the gap toward the governance- and loss-constrained trust ceiling.+20.2 pp2029 22% → now 42%

seller-agent adoptionSeller-side diffusion follows the demand agents can address and the selected annual diffusion prior.+32.6 pp2029 8% → now 40%

machine counterpartiesBuyer and seller delegation meet at a rate shaped by persistent-agent cost and branch coordination.+10.6 pp2029 2% → now 12%

continuous commerceDelegated demand, machine counterparties and governance turn isolated transactions into standing policies.+1.3 pp2029 2% → now 3%

physical executionAgent-addressable operators diffuse toward a ceiling set by seller agents, machine counterparties and governance capacity.+6.5 pp2029 0% → now 7%

pooled demandPortable pooled mandates, branch coordination and market concentration determine how much delegated demand can bargain collectively.+13.8 pp2029 0% → now 14%

adaptive supplySeller agents, negotiated transactions and physical execution let supply respond to intent before a finished item exists.+2.8 pp2029 0% → now 3%

adversarial lossFraud and manipulation pressure compete with governance capacity as machine counterparties expand.−2.3 pp2029 8% → now 5%

share of demand delegated rises to 42.1% (+20.2%); seller-agent adoption rises to 40.4% (+32.6%); share with machine counterparties rises to 12.3% (+10.6%); share negotiated rises to 6.7% (+5.0%); share run continuously rises to 3.5% (+1.3%).

New modeled actors: Merchant capability — exposes discovery, checkout, fulfillment and support as callable operations without needing a human-facing storefront; Seller agent — sets price, inventory exposure, service terms and which buyers to trust; Authority network — proves what an agent may spend, disclose, sign and delegate onward; Open commerce router — routes portable intent across catalogs, logistics and payment providers without owning the buyer relationship.

modeled from unanchored structural inputs · 17 published research hooks · ensemble ranges are modeled sensitivity, not measured probability

2036

agent markets · model-triggered authored event

Agent auctions begin clearing commercial terms.

Buyer and seller agents can form price, service and assurance terms in a federated market instead of accepting one posted bundle.

2036

Illustrative Care coordination purchase · Coordinated open rails

Terms become negotiable

Set a bargaining envelope and review only counteroffers that cross a boundary.

negotiation room · engine selected

Agent influenced$855B–$1.6T

Agent executed$398B–$935B

Merchant-routed after tolls + loss$356B–$856B

CCNegotiation roomCare coordination · authenticated workspace3 live offersCoordinated open rails2036

eligibility verifiedcontinuity requiredbooking approval only

Provider agent A$40 copay8-day intake · full fitbest evidence

Provider agent B$25 copay3-day intake · new locationreview required

Provider agent Cpendingcoverage not verifiedrejected

Buyer agent · agent auction · portable mandate · independent assurance

parallel requests for quote across a federated market. Negotiating administrative terms only; provider choice and every clinical decision remain with people.

offer window · 07:42 remainingmandate · CARE-2036-04proof attached

Illustrative outcomeThe agent binds the best policy-compliant bundle while preserving the negotiation record.

intentsourcenegotiatesettleassure

personsharedagent

Authoritybounded negotiation mandate

Market topologyfederated market

Identityportable mandate

Price formationagent auction

Settlementprogrammable escrow

Human attentionreview boundary crossings

Trust surfaceportable mandate, independent assurance, provenance and a receipt that records the mandate and result.

How the market respondsRuns parallel requests for quote, trades only the arena’s flexible terms and rejects every offer outside its mandate. Sourcing uses parallel requests for quote across a federated market; an agent auction clears the price before programmable escrow executes the commitment.

When it breakspresent a portable receipt to an independent assurer, unwind the commitment and reroute; the person’s attention mode is “review boundary crossings.”

autonomy27%

delegation fit13%

market reach72%

exception risk26%

Range-covering interface spread

approval 44%negotiate 55%policy 1%ambient 0%

512 structural draws · sensitivity coverage, not probability

Surface sensitivityThe engine-selected surface appears in 55% of the range-covering draws above. The remaining share shows which alternate purchasing interfaces the same model selects as structural assumptions move.

Evidence labelIndicators combine structural state with disclosed, unanchored arena priors. Dialogue and interface remain illustrative, not forecast outputs.

In the physical world · illustrative physical-world experience

Care logistics become continuous while clinical authority stays human.

2036 · Coordinated open rails
federated market
Delegated checkout 36%

What the person experiencesThe visible moment

A single care timeline shows appointments, eligibility, transport and benefit headroom, with a provider-change exception waiting.

What software does off-screenThe delegated work

The agent reconciles administrative commitments across clinics, benefits and mobility, and pauses whenever a clinical boundary appears.

What changes in the worldThe physical handoff

Transport, intake paperwork and coverage confirmations arrive in sequence around the person’s chosen care plan.

Control returns hereThe person and clinicians retain every diagnosis, treatment and provider decision; the agent coordinates only the surrounding work.

What happens next · model-derived interpretation

The cart stops being the center of the transaction. Buyer and seller agents exchange terms inside a mandate, while the person sees the boundary of the bargain rather than every offer. Since 2032, machine-counterparty share rises by 22.5 percentage points and continuous commerce rises by 8.1 percentage points. That state transition moves the purchasing interface from approval bundle to negotiation room.

What remains contested: A federated market expands choice, but portable mandate and independent assurance must scale without hardening into a new control point.

Active machine cast · model-derivedSeller agent · Buyer agent · Mandate agent · Settlement agent · Routing agent · Market operator · Procurement agent · Physical operator · Assurance agent · Supply agent

Coordination burden6%range 4%–9%

Human rolepolicy setter and exception handlerowns multi-agent drift

Between purchasing chapters · model-derived experience transition

The interface moves from approval bundle to negotiation room.

2032 → 2036

Interfacenegotiation roomfrom approval bundle

Human attentionreview boundary crossingsfrom approve this purchase

Agent mandatebounded negotiation mandateautonomy +12.1 pp to 27%

Market boundary72% reachablereach +0.2 pp to 72% · exception risk +0.4 pp to 26%

Model trace · why this chapter changes

Model trigger · negotiated transactions ≥ 18.0%20.5% of transactions are negotiated and seller-agent adoption reaches 71.3%.Authored consequence: Auction access and portable reputation become public-governance questions, not merely checkout features.

Event sensitivity · selected 17% · leading alternate 78%

Human approval remains the final clearing layer.5%

Negotiation protocols become shared market infrastructure.78%

Agent auctions begin clearing commercial terms.17%

512 range-covering structural draws · authored event selection coverage, not probability. Human approval remains the final clearing layer. 5% · Negotiation protocols become shared market infrastructure. 78% · Agent auctions begin clearing commercial terms. 17%

delegated demand59%

machine on both sides35%

continuous commerce12%

physical execution23%

pooled demand25%

adaptive supply11%

governance capacity76%

adversarial loss6%

2032 → 2036model-derived transition · equations, not observed attribution

delegated demandCommercial capability closes part of the gap toward the governance- and loss-constrained trust ceiling.+16.9 pp2032 42% → now 59%

seller-agent adoptionSeller-side diffusion follows the demand agents can address and the selected annual diffusion prior.+30.9 pp2032 40% → now 71%

machine counterpartiesBuyer and seller delegation meet at a rate shaped by persistent-agent cost and branch coordination.+22.5 pp2032 12% → now 35%

continuous commerceDelegated demand, machine counterparties and governance turn isolated transactions into standing policies.+8.1 pp2032 3% → now 12%

physical executionAgent-addressable operators diffuse toward a ceiling set by seller agents, machine counterparties and governance capacity.+15.8 pp2032 7% → now 23%

pooled demandPortable pooled mandates, branch coordination and market concentration determine how much delegated demand can bargain collectively.+11.4 pp2032 14% → now 25%

adaptive supplySeller agents, negotiated transactions and physical execution let supply respond to intent before a finished item exists.+8.4 pp2032 3% → now 11%

adversarial lossFraud and manipulation pressure compete with governance capacity as machine counterparties expand.+0.4 pp2032 5% → now 6%

share of demand delegated rises to 59.0% (+16.9%); seller-agent adoption rises to 71.3% (+30.9%); share with machine counterparties rises to 34.8% (+22.5%); share negotiated rises to 20.5% (+13.8%); share run continuously rises to 11.5% (+8.1%).

New modeled actors: Agent market operator — matches machine demand and supply while setting access and ranking rules; Assurance network — underwrites disputes, provenance, delivery and adversarial behavior; Physical operator agent — commits robots, vehicles, workers and facilities to machine-negotiated physical outcomes; Metered capability — quotes and settles small units of digital or physical work as they are consumed; Trust and escrow registry — makes identity, reputation, job state and conditional release portable between unfamiliar counterparties; Demand coalition — pools bounded mandates so households or organizations bargain without surrendering individual constraints.

modeled from unanchored structural inputs · 17 published research hooks · ensemble ranges are modeled sensitivity, not measured probability

2040

agent markets · model-triggered authored event

Standing policies replace more checkout sessions.

Standing policies allocate recurring demand across providers while portable reserves and receipts preserve the person’s ability to exit.

2040

Illustrative Care coordination purchase · Coordinated open rails

The purchase becomes a policy

Define the policy, review exceptions and a periodic digest, and adjust constraints.

policy console · engine selected

Agent influenced$1.2T–$2.4T

Agent executed$601B–$1.6T

Merchant-routed after tolls + loss$520B–$1.4T

CCCare coordination policyCare coordination · authenticated workspacemonitoringCoordinated open rails2040

Outcomecontinuity and access

Ceilingbenefit + $320 / month

Attention1 exception

Verify eligibility before administrative commitmenton track

Preserve provider and language continuityenforced

Any treatment or provider change needs reviewreview

Decision requested · independent assuranceApprove an earlier intake with a different provider, or keep continuity and wait five more days?standing search across a federated market · present a portable receipt to an independent assurer, unwind the commitment and reroute

policy active · exception isolatedmandate · CARE-2040-04proof attached

Illustrative outcomeA sequence of reversible commitments replaces isolated search, cart and checkout sessions.

intentsourcenegotiatesettleassure

personsharedagent

Authorityrolling category policy

Market topologyfederated market

Identityportable mandate

Price formationcontinuous clearing

Settlementrolling net settlement

Human attentionreview exceptions and digest

Trust surfaceportable mandate, independent assurance, provenance and a receipt that records the mandate and result.

How the market respondsMonitors the selected outcome, opens and closes commitments, and routes genuine exceptions to the person. Sourcing uses standing search across a federated market; continuous clearing reprices capacity before rolling net settlement executes the commitment.

When it breakspresent a portable receipt to an independent assurer, unwind the commitment and reroute; the person’s attention mode is “review exceptions and digest.”

autonomy37%

delegation fit13%

market reach72%

exception risk28%

Range-covering interface spread

approval 9%negotiate 77%policy 13%ambient 0%

512 structural draws · sensitivity coverage, not probability

Surface sensitivityThe engine-selected surface appears in 13% of the range-covering draws above. The remaining share shows which alternate purchasing interfaces the same model selects as structural assumptions move.

Evidence labelIndicators combine structural state with disclosed, unanchored arena priors. Dialogue and interface remain illustrative, not forecast outputs.

In the physical world · illustrative physical-world experience

The most capable agent is visible in the moments it refuses to act.

2040 · Coordinated open rails
federated market
Embodied orchestration 38%

What the person experiencesThe visible moment

The person sees six weeks of coordinated access and a clear pause: the next step would change provider continuity.

What software does off-screenThe delegated work

The agent keeps benefits, appointments and transport synchronized, but marks the unresolved decision instead of optimizing through it.

What changes in the worldThe physical handoff

Existing care continues, provisional capacity is held and no new provider receives authority until a person decides.

Control returns hereContinuous administrative agency never becomes continuous clinical agency.

What happens next · model-derived interpretation

The unit of commerce changes from an item to an outcome. Purchases become reversible commitments managed against a standing policy, with the person returning for exceptions and review. Since 2036, machine-counterparty share rises by 19.7 percentage points and continuous commerce rises by 11.9 percentage points. That state transition moves the purchasing interface from negotiation room to policy console.

What remains contested: A federated market expands choice, but portable mandate and independent assurance must scale without hardening into a new control point.

Active machine cast · model-derivedSeller agent · Buyer agent · Mandate agent · Settlement agent · Market operator · Routing agent · Physical operator · Assurance agent · Procurement agent · Supply agent · Object agent

Coordination burden9%range 6%–13%

Human rolepolicy setter and exception handlerowns physical recourse gap

Between purchasing chapters · model-derived experience transition

The interface moves from negotiation room to policy console.

2036 → 2040

Interfacepolicy consolefrom negotiation room

Human attentionreview exceptions and digestfrom review boundary crossings

Agent mandaterolling category policyautonomy +10.5 pp to 37%

Market boundary72% reachablereach +0.2 pp to 72% · exception risk +1.9 pp to 28%

Model trace · why this chapter changes

Model trigger · 12.0% ≤ continuous commerce < 35.0%23.5% of commerce runs continuously while market concentration is 32.6%.Authored consequence: The coordination layer must prevent assurance and routing from hardening into a new private control point.

Event sensitivity · selected 40% · leading alternate 59%

Persistent agents stay inside portable budgets.59%

Standing policies replace more checkout sessions.40%

Outcome markets begin clearing continuously.1%

512 range-covering structural draws · authored event selection coverage, not probability. Persistent agents stay inside portable budgets. 59% · Standing policies replace more checkout sessions. 40% · Outcome markets begin clearing continuously. 1%

delegated demand69%

machine on both sides54%

continuous commerce23%

physical execution38%

pooled demand33%

adaptive supply22%

governance capacity76%

adversarial loss8%

2036 → 2040model-derived transition · equations, not observed attribution

delegated demandCommercial capability closes part of the gap toward the governance- and loss-constrained trust ceiling.+9.6 pp2036 59% → now 69%

seller-agent adoptionSeller-side diffusion follows the demand agents can address and the selected annual diffusion prior.+13.9 pp2036 71% → now 85%

machine counterpartiesBuyer and seller delegation meet at a rate shaped by persistent-agent cost and branch coordination.+19.7 pp2036 35% → now 54%

continuous commerceDelegated demand, machine counterparties and governance turn isolated transactions into standing policies.+11.9 pp2036 12% → now 23%

physical executionAgent-addressable operators diffuse toward a ceiling set by seller agents, machine counterparties and governance capacity.+15.7 pp2036 23% → now 38%

pooled demandPortable pooled mandates, branch coordination and market concentration determine how much delegated demand can bargain collectively.+7.4 pp2036 25% → now 33%

adaptive supplySeller agents, negotiated transactions and physical execution let supply respond to intent before a finished item exists.+10.6 pp2036 11% → now 22%

adversarial lossFraud and manipulation pressure compete with governance capacity as machine counterparties expand.+1.9 pp2036 6% → now 8%

share of demand delegated rises to 68.6% (+9.6%); seller-agent adoption rises to 85.1% (+13.9%); share with machine counterparties rises to 54.5% (+19.7%); share negotiated rises to 36.7% (+16.1%); share run continuously rises to 23.5% (+11.9%).

New modeled actors: Resource agent — continuously allocates money, energy, compute, inventory or capacity against a policy; Asset twin — turns a machine, package, building or product lifecycle into an authenticated source of commercial intent; Production agent — configures or commissions supply after machine demand forms, before a finished product necessarily exists.

modeled from unanchored structural inputs · 17 published research hooks · ensemble ranges are modeled sensitivity, not measured probability

Expand the forces behind this futureeconomic envelope · six protocol paths · full agent ecology

Economic envelope · annual U.S. retail e-commerce flows

A dollar can be influenced, executed, retained—or captured as infrastructure.

The engine separates commerce touched by an agent from transactions actually committed by one, then estimates where value can be retained, tolled, lost or created as machine-payable work. Every range moves with this branch’s structural draws plus five disclosed economic priors.

512 paired range-covering draws
one observed 2026 anchor
not a valuation or probability

agent influencedagent executedmerchant-routed after channel value + loss

2029$1.6T
addressable baseline

Agent influenced GMV: $332B–$380Bdiscovery or decision shaped
Agent executed GMV: $121B–$187Bauthority reaches commitment
Merchant-routed value: $107B–$168Bafter modeled channel tolls + loss; not profit
Agent-channel value: $1.7B–$4.2Baccess, orchestration + distribution pool
Risk leakage: $7.0B–$19.4Bfraud, dispute, failure + exception loss
Machine-service revenue: $225M–$801Mnew metered work; may overlap other flows

2032$1.9T
addressable baseline

Agent influenced GMV: $555B–$862Bdiscovery or decision shaped
Agent executed GMV: $230B–$450Bauthority reaches commitment
Merchant-routed value: $209B–$415Bafter modeled channel tolls + loss; not profit
Agent-channel value: $3.5B–$9.7Baccess, orchestration + distribution pool
Risk leakage: $11.2B–$44.0Bfraud, dispute, failure + exception loss
Machine-service revenue: $515M–$2.2Bnew metered work; may overlap other flows

2036$2.4T
addressable baseline

Agent influenced GMV: $855B–$1.6Tdiscovery or decision shaped
Agent executed GMV: $398B–$935Bauthority reaches commitment
Merchant-routed value: $356B–$856Bafter modeled channel tolls + loss; not profit
Agent-channel value: $6.2B–$19.4Baccess, orchestration + distribution pool
Risk leakage: $15.5B–$114Bfraud, dispute, failure + exception loss
Machine-service revenue: $1.2B–$6.5Bnew metered work; may overlap other flows

2040$3.1T
addressable baseline

Agent influenced GMV: $1.2T–$2.4Tdiscovery or decision shaped
Agent executed GMV: $601B–$1.6Tauthority reaches commitment
Merchant-routed value: $520B–$1.4Tafter modeled channel tolls + loss; not profit
Agent-channel value: $9.2B–$32.0Baccess, orchestration + distribution pool
Risk leakage: $28.5B–$231Bfraud, dispute, failure + exception loss
Machine-service revenue: $2.3B–$14.5Bnew metered work; may overlap other flows

Observed anchor. $1.4T annualized from Census Q2 2026 seasonally adjusted U.S. retail e-commerce sales; future growth is not observed.26External comparison, not calibration. McKinsey estimated $900B–$1T of U.S. B2C retail revenue could be orchestrated by agentic commerce in 2030. Its definition and horizon differ from these engine flows.27 Overlapping values above must not be added.

Agentic commerce paths · 2029 → 2040

Six futures can grow from the same need.

These are paths, not end-state options. Choose a mechanism, then move through the years to see what must become true and how purchasing changes before the 2040 destination. Seller-to-seller negotiation is only one possible mechanism. Replay this care coordination need through embedded merchant capabilities, portable mandates, open commerce networks, metered services, trust markets or object-initiated coordination.

four stages per path
activation signals can coexist
experiences = illustrative

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2032
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Assistants learn to carry merchant state

Product results become structured enough for an assistant to preserve configuration, availability and intent before handing the person to a merchant.

What purchasing feels like hereThe assistant compares options and opens the exact configured merchant page; the person still reviews and completes the order there.

Conditional direct annual value pool$4.3B$1.2B–$9.3B sensitivity · based on agent-executed GMV at a 2.5% scenario take. Non-additive.

What must become true nextReliable catalog, identity and checkout state can survive the handoff.

How this path can stallMerchants block automated access or expose inconsistent product and fulfillment state.

open institutionCoordinated rules let this path cross assistants, businesses and assurance providers; the strategic contest moves to governance and routing.

Range sensitivity0% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationplatform-to-business capability

A different purchasing world · illustrative protocol-enabled purchasing experience

The assistant is the storefront

“Keep the care plan coordinated across providers and benefits while reserving all diagnosis, treatment and provider changes for a person.”
The assistant opens a live care coordination workspace. Merchant configuration, delivery choices and the final approval appear without sending the person to another site.

YouState the outcomeDescribe the care coordination result, budget and non-negotiable constraints.

AIDiscover capabilitiesIntersect the assistant profile with merchant catalog, checkout, fulfillment and identity capabilities.

BizCompose the orderThe merchant returns its own live configuration, price, delivery and support state.

YouApprove in contextOne bounded token authorizes this merchant, amount and order state.

BizFulfill and supportThe merchant stays responsible for the order, delivery, returns and relationship.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:A merchant-owned order is completed through one assistant-rendered experience; fulfillment and support remain with the business.

Mechanism dossier · conditional scenario

Businesses expose typed commerce capabilities while the assistant renders the journey and the merchant keeps the order.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

UCPACPMCP / REST

Activation signal57%

2040 direct value pool$39.4B

Dominant in range draws0%

Human surfaceone composed approval inside the assistant

Control remains hereApprove the composed order and open the merchant’s embedded surface whenever brand, fit or service judgment matters.

What could break it: A business can expose incomplete capabilities or use an embedded experience to reintroduce dark patterns.

All six mechanisms across 512 draws0% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2032
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

One-use authority becomes legible

Agent payments, identity assertions and receipts begin recording why a purchase was allowed, not only that a payment succeeded.

What purchasing feels like hereThe person approves a tightly scoped purchase token and receives a receipt that links intent, seller, amount and result.

Conditional direct annual value pool$1.2B$242M–$3.4B sensitivity · based on agent-executed GMV at a 0.7% scenario take. Non-additive.

What must become true nextWallets and merchants can represent purpose, scope, expiry and revocation consistently.

How this path can stallMandates remain unreadable legal objects or are reduced to reusable payment credentials.

open institutionCoordinated rules let this path cross assistants, businesses and assurance providers; the strategic contest moves to governance and routing.

Range sensitivity36% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationcredentialed agent-to-merchant request

A different purchasing world · illustrative protocol-enabled purchasing experience

Permission replaces checkout

“Keep the care plan coordinated across providers and benefits while reserving all diagnosis, treatment and provider changes for a person.”
The care coordination policy already contains the allowed spend, data disclosures and exception boundaries. The person sees the mandate being used—not another checkout.

YouGrant a scoped mandateAuthorize this class of care coordination outcome with explicit limits and expiry.

AIPresent minimum credentialsProve identity, eligibility and authority without disclosing the complete personal profile.

BizVerify before reservingValidate the signed intent and bind the proposed order to that authority.

NetSettle with evidencePayment and checkout receipts preserve what was authorized and what executed.

YouReview exceptions onlyRoutine compliant uses stay quiet; boundary crossings return with an explanation.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:A signed chain links the person’s authority to the exact commitment, payment and receipt.

Mechanism dossier · conditional scenario

Signed, scoped authority travels farther than any single agent, so commerce begins with a mandate and ends with auditable evidence.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

AP2Trusted AgentVerifiable Credentials

Activation signal67%

2040 direct value pool$11.0B

Dominant in range draws40%

Human surfaceauthority wallet, exceptions and receipts

Control remains hereEdit or revoke the mandate; approve only uses that cross its amount, identity, timing or consequence boundary.

What could break it: A technically valid mandate can still encode a misunderstood preference or grant more downstream delegation than intended.

All six mechanisms across 512 draws40% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2032
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Adapters translate fragmented commerce

Buyer agents normalize merchant pages and APIs into partial offer schemas, but discovery, logistics and recourse still break at network boundaries.

What purchasing feels like hereThe person receives a cross-site shortlist, then approves a bundle assembled through several visible handoffs.

Conditional direct annual value pool$2.1B$498M–$5.7B sensitivity · based on agent-influenced GMV at a 0.6% scenario take. Non-additive.

What must become true nextBasic offer, fulfillment and receipt semantics converge enough to translate.

How this path can stallEvery large participant extends the schema differently and adapters become permanent tollbooths.

open institutionCoordinated rules let this path cross assistants, businesses and assurance providers; the strategic contest moves to governance and routing.

Range sensitivity64% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationfederated demand-and-supply network

A different purchasing world · illustrative protocol-enabled purchasing experience

Commerce becomes a routing network

“Keep the care plan coordinated across providers and benefits while reserving all diagnosis, treatment and provider changes for a person.”
One care coordination request fans out across independent buyer, seller, logistics and assurance participants, then returns as a composed outcome.

YouPublish bounded intentSend a time-limited care coordination request without selecting a marketplace first.

NetRoute across networksGateways find compatible seller, logistics, financing and assurance participants.

BizReturn interoperable offersParticipants answer the same schema with inventory, terms, evidence and service capacity.

AICompose the outcomeRank whole outcomes rather than isolated products and preserve independent alternatives.

NetCoordinate fulfillmentCommitments and status updates travel across the selected participant graph.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:The selected bundle can combine inventory, service and physical execution from participants that never shared one marketplace owner.

Mechanism dossier · conditional scenario

Demand is broadcast into interoperable buyer, seller, logistics and assurance networks instead of sent to one seller agent.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

BecknONDCUCP discovery

Activation signal59%

2040 direct value pool$12.7B

Dominant in range draws60%

Human surfacea composed outcome from many network participants

Control remains hereChoose the routing and ranking policy, inspect participant provenance and exclude any network or seller class.

What could break it: Fragmented governance can make responsibility ambiguous when several participants jointly create a failed outcome.

All six mechanisms across 512 draws60% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2040
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Agents buy small pieces of digital work

Machine payments make data, search, verification and generation callable without a subscription or procurement cycle.

What purchasing feels like hereThe person approves a capped task; the assistant buys a few services and shows cost beside the resulting recommendation.

Conditional direct annual value pool$483M$225M–$801M sensitivity · based on metered machine services. Non-additive.

What must become true nextCapabilities can quote, authenticate, settle and return useful evidence cheaply.

How this path can stallMicro-payments add friction without proving that the work improved the outcome.

open institutionCoordinated rules let this path cross assistants, businesses and assurance providers; the strategic contest moves to governance and routing.

Range sensitivity0% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationmachine-payable service graph

A different purchasing world · illustrative protocol-enabled purchasing experience

Everything becomes a paid capability

“Keep the care plan coordinated across providers and benefits while reserving all diagnosis, treatment and provider changes for a person.”
Instead of purchasing a finished care coordination package, the assistant opens paid sessions for the data, capacity and execution needed right now.

YouFund an outcome reservePlace a capped reserve behind the care coordination policy rather than prepaying one bundle.

AIDiscover paid capabilitiesFind machine-readable services for evidence, planning, capacity and execution.

NetOpen metered sessionsAuthorize capped sessions and stream settlement as useful work is verified.

AIRecompose continuouslyReplace underperforming services without discarding the completed work or policy.

YouAudit spend and resultSee cost by capability, final evidence and every automatically stopped session.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:The reserve is charged as verified work arrives, allowing the service graph to change without rebuilding a cart.

Mechanism dossier · conditional scenario

The commercial unit becomes a session, result or stream of work rather than a cart containing a finished product.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

MPPx402MCP

Activation signal27%

2040 direct value pool$13.4B

Dominant in range draws0%

Human surfacea reserve, live spend stream and result ledger

Control remains hereSet the reserve, per-service ceilings and stop conditions; review the live spend stream or terminate every session.

What could break it: Many individually cheap calls can create runaway spend, dependency chains or incentives to prolong the task.

All six mechanisms across 512 draws0% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2036
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Agents inherit platform reputation

Assistants can use marketplace ratings and escrow, but identity, ranking and dispute history remain trapped inside each platform.

What purchasing feels like hereThe person sees which platform guarantee backs the recommendation and must enter that platform to resolve failure.

Conditional direct annual value pool$2.1B$363M–$5.6B sensitivity · based on agent-executed GMV at a 1.2% scenario take. Non-additive.

What must become true nextAgent actions can be attributed to verified accounts and bounded payment instruments.

How this path can stallSynthetic reviews and account farming make inherited reputation less useful to machines.

open institutionCoordinated rules let this path cross assistants, businesses and assurance providers; the strategic contest moves to governance and routing.

Range sensitivity0% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationregistry-and-escrow work market

A different purchasing world · illustrative protocol-enabled purchasing experience

Trust becomes the marketplace

“Keep the care plan coordinated across providers and benefits while reserving all diagnosis, treatment and provider changes for a person.”
The care coordination request becomes a temporary market: providers bond a claim, independent evaluators inspect the result and escrow releases against evidence.

AIPost a verifiable jobPublish the care coordination result, acceptance tests, deadline and escrow terms.

NetFilter identity and reputationResolve registered providers, attestations, prior outcomes and required bonds.

BizBid with evidenceProviders commit price, method, service terms and a bond against non-performance.

NetEvaluate the outcomeIndependent evaluators check delivery evidence and open a dispute window.

YouAccept or appealEscrow releases automatically inside policy; contested judgments return with the full record.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:Unknown providers can compete without requiring one platform to own identity, ranking, payment and dispute resolution.

Mechanism dossier · conditional scenario

Identity, reputation, escrow and independent evaluation let unknown agents form temporary markets around a job.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

ERC-8004ERC-8183A2A

Activation signal57%

2040 direct value pool$18.9B

Dominant in range draws0%

Human surfaceproof thresholds, bonds, evaluators and recourse

Control remains hereChoose reputation sources, minimum bonds, evaluators and appeal rules before the job is posted.

What could break it: Reputation can be gamed, evaluators can collude and on-chain evidence can preserve a wrong judgment permanently.

All six mechanisms across 512 draws0% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2040
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Products expose a persistent identity

Digital links and event histories let a person or assistant retrieve authoritative product, ownership and service information after purchase.

What purchasing feels like hereThe person scans or asks the assistant about the asset; service history and compatible options replace a fresh web search.

Conditional direct annual value pool$160M$39.8M–$399M sensitivity · based on physically orchestrated GMV at a 3.0% scenario take. Non-additive.

What must become true nextProduct identifiers resolve to portable, owner-accessible lifecycle data.

How this path can stallManufacturers control the history, withhold repair data or abandon the resolver.

open institutionCoordinated rules let this path cross assistants, businesses and assurance providers; the strategic contest moves to governance and routing.

Range sensitivity0% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationasset-to-service coordination loop

A different purchasing world · illustrative protocol-enabled purchasing experience

Physical objects become the buyers

“Keep the care plan coordinated across providers and benefits while reserving all diagnosis, treatment and provider changes for a person.”
The physical asset notices the care coordination condition first. Its live history explains what changed, which policy applies and what response is being prepared.

ObjDetect a state changeThe asset records the care coordination condition with time, location, provenance and sensor evidence.

ObjCheck its service policyDetermine whether to observe, reserve capacity, request approval or stop operation.

NetFind compatible physical serviceQuery parts, technicians, facilities and logistics against the asset identity and history.

AIAuthorize the responseApply the owner’s mandate, preserve alternatives and bind the selected physical plan.

ObjVerify the new stateRecord completion evidence, update the twin and release payment only for the verified result.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:The asset’s condition, service history and verified result form one continuing commercial record instead of separate shopping and maintenance journeys.

Mechanism dossier · conditional scenario

Assets publish their own condition and service history, allowing a vehicle, appliance, building or inventory item to initiate commerce.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

GS1 Digital LinkEPCISdigital twins

Activation signal32%

2040 direct value pool$35.4B

Dominant in range draws0%

Human surfacea living asset timeline with revocable operating policy

Control remains hereControl which sensors and service classes may initiate action, require approval for irreversible work and revoke the asset mandate.

What could break it: Bad sensor data, compromised twins or manufacturer-controlled histories can trigger unnecessary work or lock owners into captive service.

All six mechanisms across 512 draws0% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From one assistant to an agent ecology

Eight forms can coexist—and diverge.

These are independent activation signals from the same structural state, not stages, market shares or probabilities. A seller site can persist while pooled demand, commissioned supply or physical orchestration grows beside it.

strongest form outlined
physical implications authored

interaction form2029203220362040

Site referralPhysical fulfillment remains downstream of a conventional order.

77%

51%

27%

14%

Delegated checkoutThe seller fulfills an agent-authored order rather than a person-built cart.

20%

35%

36%

29%

Bilateral agent tradeInventory, service capacity and delivery can be reserved machine-to-machine.

24%

32%

Terms marketThe physical plan changes before checkout because timing and capacity are part of the bargain.

17%

32%

Standing outcome policySupply, repair, transport or care coordination continue without a new shopping session.

12%

23%

Collective procurementNeighborhoods, fleets or firms reshape routes, inventory and capacity through coordinated demand.

14%

25%

33%

Commissioned supplyProduction schedules, components and facilities respond to intent rather than forecasted shelf demand.

11%

22%

Embodied orchestrationThe purchasing interface becomes a live map of actions, safety boundaries and revocable authority.

23%

38%

Delegation frontier · published path

Where autonomy stops depends on the need.

The aggregate market state is identical down each column. Only the disclosed arena priors change. Each cell reports its preset surface and the full interface spread across 512 range-covering structural draws; segment widths are sensitivity coverage, not probability.

engine-selected interface + full range coverage · unanchored priors · illustrative UI

commerce arena2029203220362040

Household replenishmentexplore this needapproval100% selectedpolicy64% selectedambient56% selectedambient91% selected

Everyday mobilityexplore this needapproval100% selectednegotiate64% selectedpolicy55% selectedambient13% selected

Complex travelexplore this needapproval100% selectedapproval99% selectednegotiate55% selectedpolicy13% selected

Care coordinationselected lensapproval100% selectedapproval99% selectednegotiate55% selectedpolicy13% selected

approvalnegotiatepolicyambient

Published path · structural model

Fragmented open web

No runtime fully closes the door, but no coordination layer makes every buyer instrument and seller rail work together either. Merchants repeat integrations, agents maintain adapters and cross-rail conversion remains partial. By 2029Q2 agents carry 19% of online GMV and 21% of attempts settle — better than the gardens, but three points of GMV and eight points of wallet adoption behind coordinated open rails.

2040 machine counterparties22%

continuous commerce6%

published 2040 p10–p9012%–34%512 structural draws

The whole path in four changes

One need moves from request to operating policy.

story first · model layers after

2029 · assisted commerceIntent becomes a mandate

The person gives software a bounded right to act.

purchase: approval bundle
agent-executed: $102B–$158B
machine cast: 3 role classes
human: operator

2032 · delegated commerceBusinesses become callable

Catalog, checkout, fulfillment and support become machine-addressable capabilities.

purchase: approval bundle
agent-executed: $192B–$379B
machine cast: 5 role classes
human: operator

2036 · delegated commerceThe market gains counterparties

Specialized buyer, seller, trust and operating agents coordinate the purchase.

purchase: approval bundle
agent-executed: $327B–$772B
machine cast: 8 role classes
human: approver and recourse owner

2040 · delegated commerceThe purchase becomes a policy

Standing intent allocates money and physical capacity until a boundary is crossed.

purchase: negotiation room
agent-executed: $491B–$1.3T
machine cast: 10 role classes
human: approver and recourse owner

Conditional annual ranges overlap and are not valuations. Role classes are functional activations above the disclosed threshold—not agent population or forecast headcount.

2029

assisted commerce · model-triggered authored event

Open protocols produce an adapter economy.

No universal coordination layer wins, so agents and merchants rely on adapters to translate instruments, terms and receipts.

2029

Illustrative Care coordination purchase · Fragmented open web

One intent, one approval

State the goal, inspect the evidence and approve this purchase.

approval bundle · engine selected

Agent influenced$298B–$341B

Agent executed$102B–$158B

Merchant-routed after tolls + loss$88.2B–$139B

CCAuthorization bundleCare coordination · authenticated workspaceready for approvalFragmented open web2029

IntentFind three eligible care providers with the right specialty, availability and language fit; do not book without approval.

Recommended bundleEligible providers · language fit · verified availabilityCoverage, credentials and continuity evidence remain inspectable · translated credentials · multi-party evidence

within mandate

FallbackEarlier intake · different locationNever booked without explicit provider approval · reconcile signed adapter logs, pause the disputed leg and ask for a bounded human decision

person decides

One signed commitment · one-shot settlementAuthorize purchase

terms verifiedmandate · CARE-2029-04proof attached

Illustrative outcomeThe person approves one bounded commitment and receives a receipt plus execution watch.

intentsourcenegotiatesettleassure

personsharedagent

Authoritysingle-use mandate

Market topologyadapter mesh

Identitytranslated credentials

Price formationposted bundle

Settlementone-shot settlement

Human attentionapprove this purchase

Trust surfacetranslated credentials, multi-party evidence, provenance and a receipt that records the mandate and result.

How the market respondsAssembles a ranked offer, counterparty, settlement and assurance bundle, then waits for one explicit authorization. Sourcing uses ranked shortlist from an adapter mesh; posted bundles set the price before one-shot settlement executes the commitment.

When it breaksreconcile signed adapter logs, pause the disputed leg and ask for a bounded human decision; the person’s attention mode is “approve this purchase.”

autonomy6%

delegation fit13%

market reach60%

exception risk35%

Range-covering interface spread

approval 100%negotiate 0%policy 0%ambient 0%

512 structural draws · sensitivity coverage, not probability

Evidence labelIndicators combine structural state with disclosed, unanchored arena priors. Dialogue and interface remain illustrative, not forecast outputs.

In the physical world · illustrative physical-world experience

The agent prepares the visit; the person chooses the provider.

2029 · Fragmented open web
adapter mesh
Site referral 79%

What the person experiencesThe visible moment

The person sees three eligible providers with verified specialty, language fit, availability and coverage evidence.

What software does off-screenThe delegated work

The agent verifies administrative facts, resolves contradictory directories and prepares—but does not submit—the booking.

What changes in the worldThe physical handoff

After explicit approval, the clinic receives the minimum eligibility and scheduling information needed for intake.

Control returns hereProvider selection, booking and every clinical choice remain with a person.

What happens next · model-derived interpretation

The visible change is compression: product search, seller choice, service, payment and delivery arrive as one proposal. Human attention concentrates at the moment of authorization. At the 2029 handoff, 20% of demand is delegated while only 2% has a machine on both sides. The experience therefore begins as an approval bundle.

What remains contested: The adapter mesh preserves contestability, but its translated credentials model makes assurance and exception recovery harder to reconcile across providers.

Active machine cast · model-derivedBuyer agent · Mandate agent · Settlement agent

Coordination burden3%range 2%–3%

Human roleoperatorowns multi-agent drift

Between purchasing chapters · model-derived experience transition

The structural handoff begins as an approval bundle.

2029Q2 → 2029

Interfaceapproval bundlestructural starting state

Human attentionapprove this purchasestructural starting state

Agent mandatesingle-use mandateautonomy starts at 6%

Market boundary60% reachablereach starts at 60% · exception risk starts at 35%

Model trace · why this chapter changes

Model trigger · branch-specific event at the near-term handoff19.7% of demand is delegated, while seller-agent adoption begins at 7.8% after the 2029Q2 handoff.Authored consequence: Compatibility becomes a competitive service whose errors are discovered during reconciliation and recovery.

Event sensitivity · branch-fixed handoff event

Open protocols produce an adapter economy.100%

512 range-covering structural draws · authored event selection coverage, not probability. Open protocols produce an adapter economy. 100%

delegated demand20%

machine on both sides2%

continuous commerce2%

physical execution0%

pooled demand0%

adaptive supply0%

governance capacity49%

adversarial loss8%

2029Q2 → structural 2029model-derived transition · equations, not observed attribution

delegated demandThe structural layer carries the branch’s final agent-GMV share into its starting delegated-demand state.20%near-term agent GMV 20% → 20%

seller-agent footholdA disclosed 12% handoff factor converts rail-capable sellers into the initial seller-agent foothold.8%sellers answering a rail 65% → 8%

governance capacityPortable authority, liability clarity and branch coordination are averaged into the initial governance capacity.49%structural priors

physical execution footholdA disclosed handoff factor seeds the small share of agent demand that can already cause and verify physical execution.0%near-term agent GMV + structural prior 20% → 0%

pooled demand footholdDelegated demand and the pooled-mandate prior seed the initial ability to bargain across more than one buyer.0%agent GMV + pooled-mandate prior 20% → 0%

adaptive supply footholdA disclosed handoff factor seeds supply that can be configured from machine-readable demand rather than selected from a fixed catalog.0%seller-agent foothold

adversarial lossThe failed-attempt share and adversarial-pressure prior seed loss pressure; this is a structural mapping, not observed loss.8%failed attempts + pressure prior 78% → 8%

share of demand delegated starts at 19.7%; seller-agent adoption starts at 7.8%; share with machine counterparties starts at 1.5%; share negotiated starts at 1.6%; share run continuously starts at 2.0%.

No new actor class enters yet; adoption and institutional state drive this transition.

modeled from unanchored structural inputs · 17 published research hooks · ensemble ranges are modeled sensitivity, not measured probability

2032

delegated commerce · model-triggered authored event

Credential translation becomes its own market.

Translation providers assemble partial authority and reputation across incompatible runtimes, sellers and payment rails.

2032

Illustrative Care coordination purchase · Fragmented open web

One intent, one approval

State the goal, inspect the evidence and approve this purchase.

approval bundle · engine selected

Agent influenced$496B–$778B

Agent executed$192B–$379B

Merchant-routed after tolls + loss$170B–$340B

CCAuthorization bundleCare coordination · authenticated workspaceready for approvalFragmented open web2032

IntentNegotiate intake timing and verified coverage while preserving provider choice and continuity constraints.

Recommended bundleEligible providers · language fit · verified availabilityCoverage, credentials and continuity evidence remain inspectable · translated credentials · multi-party evidence

within mandate

FallbackEarlier intake · different locationNever booked without explicit provider approval · reconcile signed adapter logs, pause the disputed leg and ask for a bounded human decision

person decides

One signed commitment · one-shot settlementAuthorize purchase

terms verifiedmandate · CARE-2032-04proof attached

Illustrative outcomeThe person approves one bounded commitment and receives a receipt plus execution watch.

intentsourcenegotiatesettleassure

personsharedagent

Authoritysingle-use mandate

Market topologyadapter mesh

Identitytranslated credentials

Price formationposted bundle

Settlementone-shot settlement

Human attentionapprove this purchase

Trust surfacetranslated credentials, multi-party evidence, provenance and a receipt that records the mandate and result.

How the market respondsAssembles a ranked offer, counterparty, settlement and assurance bundle, then waits for one explicit authorization. Sourcing uses ranked shortlist from an adapter mesh; posted bundles set the price before one-shot settlement executes the commitment.

When it breaksreconcile signed adapter logs, pause the disputed leg and ask for a bounded human decision; the person’s attention mode is “approve this purchase.”

autonomy10%

delegation fit13%

market reach60%

exception risk34%

Range-covering interface spread

approval 100%negotiate 0%policy 0%ambient 0%

512 structural draws · sensitivity coverage, not probability

Evidence labelIndicators combine structural state with disclosed, unanchored arena priors. Dialogue and interface remain illustrative, not forecast outputs.

In the physical world · illustrative physical-world experience

Administrative terms negotiate; care decisions do not.

2032 · Fragmented open web
adapter mesh
Site referral 64%

What the person experiencesThe visible moment

The person sees an earlier intake at a new location beside a later appointment that preserves continuity and language fit.

What software does off-screenThe delegated work

Provider, benefits and transport agents negotiate eligibility, intake timing and logistics without exchanging clinical discretion.

What changes in the worldThe physical handoff

The chosen clinic, ride and benefits authorization are coordinated only after the person selects the provider path.

Control returns hereThe mandate excludes diagnosis, treatment and autonomous provider substitution.

What happens next · model-derived interpretation

The visible change is compression: product search, seller choice, service, payment and delivery arrive as one proposal. Human attention concentrates at the moment of authorization. Since 2029, machine-counterparty share rises by 3.8 percentage points and continuous commerce falls by 0.3 percentage points. The experience remains an approval bundle, but its mandate and market scope deepen.

What remains contested: The adapter mesh preserves contestability, but its translated credentials model makes assurance and exception recovery harder to reconcile across providers.

Active machine cast · model-derivedBuyer agent · Seller agent · Mandate agent · Settlement agent · Routing agent

Coordination burden3%range 2%–4%

Human roleoperatorowns multi-agent drift

Between purchasing chapters · model-derived experience transition

approval bundle remains visible while its mandate deepens.

2029 → 2032

Interfaceapproval bundleunchanged from approval bundle

Human attentionapprove this purchaseunchanged from approve this purchase

Agent mandatesingle-use mandateautonomy +4.0 pp to 10%

Market boundary60% reachablereach +0.0 pp to 60% · exception risk -0.9 pp to 34%

Model trace · why this chapter changes

Model trigger · 35.0% ≤ identity portability < 65.0%Identity portability is 46.0% and machine counterparties reach 5.3%.Authored consequence: The market stays contestable, but buyers must decide which translator can be trusted with accumulated commercial context.

Event sensitivity · selected 38% · leading alternate 31%

Adapters inherit the authority problem.31%

Credential translation becomes its own market.38%

A portability compact breaks the adapter bottleneck.31%

512 range-covering structural draws · authored event selection coverage, not probability. Adapters inherit the authority problem. 31% · Credential translation becomes its own market. 38% · A portability compact breaks the adapter bottleneck. 31%

delegated demand32%

machine on both sides5%

continuous commerce2%

physical execution3%

pooled demand5%

adaptive supply2%

governance capacity49%

adversarial loss7%

2029 → 2032model-derived transition · equations, not observed attribution

delegated demandCommercial capability closes part of the gap toward the governance- and loss-constrained trust ceiling.+12.4 pp2029 20% → now 32%

seller-agent adoptionSeller-side diffusion follows the demand agents can address and the selected annual diffusion prior.+17.7 pp2029 8% → now 25%

machine counterpartiesBuyer and seller delegation meet at a rate shaped by persistent-agent cost and branch coordination.+3.8 pp2029 2% → now 5%

continuous commerceDelegated demand, machine counterparties and governance turn isolated transactions into standing policies.−0.3 pp2029 2% → now 2%

physical executionAgent-addressable operators diffuse toward a ceiling set by seller agents, machine counterparties and governance capacity.+2.7 pp2029 0% → now 3%

pooled demandPortable pooled mandates, branch coordination and market concentration determine how much delegated demand can bargain collectively.+4.5 pp2029 0% → now 5%

adaptive supplySeller agents, negotiated transactions and physical execution let supply respond to intent before a finished item exists.+1.4 pp2029 0% → now 2%

adversarial lossFraud and manipulation pressure compete with governance capacity as machine counterparties expand.−0.9 pp2029 8% → now 7%

share of demand delegated rises to 32.1% (+12.4%); seller-agent adoption rises to 25.4% (+17.7%); share with machine counterparties rises to 5.3% (+3.8%); share negotiated rises to 2.9% (+1.3%); share run continuously falls to 1.7% (-0.3%).

modeled from unanchored structural inputs · 17 published research hooks · ensemble ranges are modeled sensitivity, not measured probability

2036

delegated commerce · model-triggered authored event

Seller agents become negotiating counterparties.

Seller agents negotiate through overlapping adapter networks, creating a market for normalization, evidence and dispute routing.

2036

Illustrative Care coordination purchase · Fragmented open web

One intent, one approval

State the goal, inspect the evidence and approve this purchase.

approval bundle · engine selected

Agent influenced$758B–$1.4T

Agent executed$327B–$772B

Merchant-routed after tolls + loss$284B–$698B

CCAuthorization bundleCare coordination · authenticated workspaceready for approvalFragmented open web2036

IntentMaintain care continuity below the benefit ceiling; coordinate eligibility, appointments and transport, escalating every clinical choice.

Recommended bundleEligible providers · language fit · verified availabilityCoverage, credentials and continuity evidence remain inspectable · translated credentials · multi-party evidence

within mandate

FallbackEarlier intake · different locationNever booked without explicit provider approval · reconcile signed adapter logs, pause the disputed leg and ask for a bounded human decision

person decides

One signed commitment · conditional atomic settlementAuthorize purchase

terms verifiedmandate · CARE-2036-04proof attached

Illustrative outcomeThe person approves one bounded commitment and receives a receipt plus execution watch.

intentsourcenegotiatesettleassure

personsharedagent

Authoritysingle-use mandate

Market topologyadapter mesh

Identitytranslated credentials

Price formationnegotiated bundle

Settlementconditional atomic settlement

Human attentionapprove this purchase

Trust surfacetranslated credentials, multi-party evidence, provenance and a receipt that records the mandate and result.

How the market respondsAssembles a ranked offer, counterparty, settlement and assurance bundle, then waits for one explicit authorization. Sourcing uses ranked shortlist from an adapter mesh; buyer and seller agents negotiate the bundle before conditional atomic settlement executes the commitment.

When it breaksreconcile signed adapter logs, pause the disputed leg and ask for a bounded human decision; the person’s attention mode is “approve this purchase.”

autonomy15%

delegation fit13%

market reach60%

exception risk34%

Range-covering interface spread

approval 76%negotiate 24%policy 0%ambient 0%

512 structural draws · sensitivity coverage, not probability

Surface sensitivityThe engine-selected surface appears in 76% of the range-covering draws above. The remaining share shows which alternate purchasing interfaces the same model selects as structural assumptions move.

Evidence labelIndicators combine structural state with disclosed, unanchored arena priors. Dialogue and interface remain illustrative, not forecast outputs.

In the physical world · illustrative physical-world experience

Care logistics become continuous while clinical authority stays human.

2036 · Fragmented open web
adapter mesh
Site referral 49%

What the person experiencesThe visible moment

A single care timeline shows appointments, eligibility, transport and benefit headroom, with a provider-change exception waiting.

What software does off-screenThe delegated work

The agent reconciles administrative commitments across clinics, benefits and mobility, and pauses whenever a clinical boundary appears.

What changes in the worldThe physical handoff

Transport, intake paperwork and coverage confirmations arrive in sequence around the person’s chosen care plan.

Control returns hereThe person and clinicians retain every diagnosis, treatment and provider decision; the agent coordinates only the surrounding work.

What happens next · model-derived interpretation

The visible change is compression: product search, seller choice, service, payment and delivery arrive as one proposal. Human attention concentrates at the moment of authorization. Since 2032, machine-counterparty share rises by 8.5 percentage points and continuous commerce rises by 1.7 percentage points. The experience remains an approval bundle, but its mandate and market scope deepen.

What remains contested: The adapter mesh preserves contestability, but its translated credentials model makes assurance and exception recovery harder to reconcile across providers.

Active machine cast · model-derivedSeller agent · Buyer agent · Mandate agent · Settlement agent · Routing agent · Market operator · Physical operator · Procurement agent

Coordination burden5%range 4%–8%

Human roleapprover and recourse ownerowns multi-agent drift

Between purchasing chapters · model-derived experience transition

approval bundle remains visible while its mandate deepens.

2032 → 2036

Interfaceapproval bundleunchanged from approval bundle

Human attentionapprove this purchaseunchanged from approve this purchase

Agent mandatesingle-use mandateautonomy +5.4 pp to 15%

Market boundary60% reachablereach +0.0 pp to 60% · exception risk +0.3 pp to 34%

Model trace · why this chapter changes

Model trigger · 6.0% ≤ negotiated transactions < 18.0%7.1% of transactions are negotiated and seller-agent adoption reaches 45.1%.Authored consequence: A deal can clear without a dominant platform, yet no single operator owns the complete evidence needed to unwind it.

Event sensitivity · selected 68% · leading alternate 32%

Translation cost caps machine negotiation.32%

Seller agents become negotiating counterparties.68%

Competing adapters converge on auction semantics.0%

512 range-covering structural draws · authored event selection coverage, not probability. Translation cost caps machine negotiation. 32% · Seller agents become negotiating counterparties. 68% · Competing adapters converge on auction semantics. 0%

delegated demand43%

machine on both sides14%

continuous commerce3%

physical execution9%

pooled demand8%

adaptive supply5%

governance capacity49%

adversarial loss8%

2032 → 2036model-derived transition · equations, not observed attribution

delegated demandCommercial capability closes part of the gap toward the governance- and loss-constrained trust ceiling.+11.1 pp2032 32% → now 43%

seller-agent adoptionSeller-side diffusion follows the demand agents can address and the selected annual diffusion prior.+19.6 pp2032 25% → now 45%

machine counterpartiesBuyer and seller delegation meet at a rate shaped by persistent-agent cost and branch coordination.+8.5 pp2032 5% → now 14%

continuous commerceDelegated demand, machine counterparties and governance turn isolated transactions into standing policies.+1.7 pp2032 2% → now 3%

physical executionAgent-addressable operators diffuse toward a ceiling set by seller agents, machine counterparties and governance capacity.+6.5 pp2032 3% → now 9%

pooled demandPortable pooled mandates, branch coordination and market concentration determine how much delegated demand can bargain collectively.+3.8 pp2032 5% → now 8%

adaptive supplySeller agents, negotiated transactions and physical execution let supply respond to intent before a finished item exists.+3.6 pp2032 2% → now 5%

adversarial lossFraud and manipulation pressure compete with governance capacity as machine counterparties expand.+0.3 pp2032 7% → now 8%

share of demand delegated rises to 43.1% (+11.1%); seller-agent adoption rises to 45.1% (+19.6%); share with machine counterparties rises to 13.8% (+8.5%); share negotiated rises to 7.1% (+4.2%); share run continuously rises to 3.4% (+1.7%).

modeled from unanchored structural inputs · 17 published research hooks · ensemble ranges are modeled sensitivity, not measured probability

2040

delegated commerce · model-triggered authored event

Persistent agents remain bounded by reconciliation debt.

Persistent agents compose standing services across an open web whose credentials, receipts and recovery paths remain only partly aligned.

2040

Illustrative Care coordination purchase · Fragmented open web

Terms become negotiable

Set a bargaining envelope and review only counteroffers that cross a boundary.

negotiation room · engine selected

Agent influenced$1.0T–$2.2T

Agent executed$491B–$1.3T

Merchant-routed after tolls + loss$413B–$1.1T

CCNegotiation roomCare coordination · authenticated workspace3 live offersFragmented open web2040

eligibility verifiedcontinuity requiredbooking approval only

Provider agent A$40 copay8-day intake · full fitbest evidence

Provider agent B$25 copay3-day intake · new locationreview required

Provider agent Cpendingcoverage not verifiedrejected

Buyer agent · negotiated bundle · translated credentials · multi-party evidence

parallel requests for quote across an adapter mesh. Negotiating administrative terms only; provider choice and every clinical decision remain with people.

offer window · 07:42 remainingmandate · CARE-2040-04proof attached

Illustrative outcomeThe agent binds the best policy-compliant bundle while preserving the negotiation record.

intentsourcenegotiatesettleassure

personsharedagent

Authoritybounded negotiation mandate

Market topologyadapter mesh

Identitytranslated credentials

Price formationnegotiated bundle

Settlementprogrammable escrow

Human attentionreview boundary crossings

Trust surfacetranslated credentials, multi-party evidence, provenance and a receipt that records the mandate and result.

How the market respondsRuns parallel requests for quote, trades only the arena’s flexible terms and rejects every offer outside its mandate. Sourcing uses parallel requests for quote across an adapter mesh; buyer and seller agents negotiate the bundle before programmable escrow executes the commitment.

When it breaksreconcile signed adapter logs, pause the disputed leg and ask for a bounded human decision; the person’s attention mode is “review boundary crossings.”

autonomy20%

delegation fit13%

market reach60%

exception risk35%

Range-covering interface spread

approval 33%negotiate 66%policy 1%ambient 0%

512 structural draws · sensitivity coverage, not probability

Surface sensitivityThe engine-selected surface appears in 66% of the range-covering draws above. The remaining share shows which alternate purchasing interfaces the same model selects as structural assumptions move.

Evidence labelIndicators combine structural state with disclosed, unanchored arena priors. Dialogue and interface remain illustrative, not forecast outputs.

In the physical world · illustrative physical-world experience

The most capable agent is visible in the moments it refuses to act.

2040 · Fragmented open web
adapter mesh
Site referral 39%

What the person experiencesThe visible moment

The person sees six weeks of coordinated access and a clear pause: the next step would change provider continuity.

What software does off-screenThe delegated work

The agent keeps benefits, appointments and transport synchronized, but marks the unresolved decision instead of optimizing through it.

What changes in the worldThe physical handoff

Existing care continues, provisional capacity is held and no new provider receives authority until a person decides.

Control returns hereContinuous administrative agency never becomes continuous clinical agency.

What happens next · model-derived interpretation

The cart stops being the center of the transaction. Buyer and seller agents exchange terms inside a mandate, while the person sees the boundary of the bargain rather than every offer. Since 2036, machine-counterparty share rises by 8.6 percentage points and continuous commerce rises by 3.1 percentage points. That state transition moves the purchasing interface from approval bundle to negotiation room.

What remains contested: The adapter mesh preserves contestability, but its translated credentials model makes assurance and exception recovery harder to reconcile across providers.

Active machine cast · model-derivedSeller agent · Buyer agent · Mandate agent · Settlement agent · Routing agent · Market operator · Physical operator · Assurance agent · Procurement agent · Supply agent

Coordination burden8%range 5%–12%

Human roleapprover and recourse ownerowns multi-agent drift

Between purchasing chapters · model-derived experience transition

The interface moves from approval bundle to negotiation room.

2036 → 2040

Interfacenegotiation roomfrom approval bundle

Human attentionreview boundary crossingsfrom approve this purchase

Agent mandatebounded negotiation mandateautonomy +4.8 pp to 20%

Market boundary60% reachablereach +0.0 pp to 60% · exception risk +1.2 pp to 35%

Model trace · why this chapter changes

Model trigger · continuous commerce < 12.0%6.4% of commerce runs continuously while market concentration is 46.9%.Authored consequence: Continuity depends on translators staying solvent, interoperable and accountable long after the original purchase.

Event sensitivity · selected 89% · leading alternate 11%

Persistent agents remain bounded by reconciliation debt.89%

Policy brokers assemble continuous service across the open web.11%

Continuous clearing emerges across adapter federations.0%

512 range-covering structural draws · authored event selection coverage, not probability. Persistent agents remain bounded by reconciliation debt. 89% · Policy brokers assemble continuous service across the open web. 11% · Continuous clearing emerges across adapter federations. 0%

delegated demand50%

machine on both sides22%

continuous commerce6%

physical execution16%

pooled demand11%

adaptive supply10%

governance capacity49%

adversarial loss9%

2036 → 2040model-derived transition · equations, not observed attribution

delegated demandCommercial capability closes part of the gap toward the governance- and loss-constrained trust ceiling.+7.0 pp2036 43% → now 50%

seller-agent adoptionSeller-side diffusion follows the demand agents can address and the selected annual diffusion prior.+12.1 pp2036 45% → now 57%

machine counterpartiesBuyer and seller delegation meet at a rate shaped by persistent-agent cost and branch coordination.+8.6 pp2036 14% → now 22%

continuous commerceDelegated demand, machine counterparties and governance turn isolated transactions into standing policies.+3.1 pp2036 3% → now 6%

physical executionAgent-addressable operators diffuse toward a ceiling set by seller agents, machine counterparties and governance capacity.+7.0 pp2036 9% → now 16%

pooled demandPortable pooled mandates, branch coordination and market concentration determine how much delegated demand can bargain collectively.+2.6 pp2036 8% → now 11%

adaptive supplySeller agents, negotiated transactions and physical execution let supply respond to intent before a finished item exists.+4.4 pp2036 5% → now 10%

adversarial lossFraud and manipulation pressure compete with governance capacity as machine counterparties expand.+1.2 pp2036 8% → now 9%

share of demand delegated rises to 50.2% (+7.0%); seller-agent adoption rises to 57.1% (+12.1%); share with machine counterparties rises to 22.4% (+8.6%); share negotiated rises to 12.6% (+5.5%); share run continuously rises to 6.4% (+3.1%).

modeled from unanchored structural inputs · 17 published research hooks · ensemble ranges are modeled sensitivity, not measured probability

Expand the forces behind this futureeconomic envelope · six protocol paths · full agent ecology

Economic envelope · annual U.S. retail e-commerce flows

A dollar can be influenced, executed, retained—or captured as infrastructure.

512 paired range-covering draws
one observed 2026 anchor
not a valuation or probability

agent influencedagent executedmerchant-routed after channel value + loss

2029$1.6T
addressable baseline

Agent influenced GMV: $298B–$341Bdiscovery or decision shaped
Agent executed GMV: $102B–$158Bauthority reaches commitment
Merchant-routed value: $88.2B–$139Bafter modeled channel tolls + loss; not profit
Agent-channel value: $3.0B–$7.5Baccess, orchestration + distribution pool
Risk leakage: $6.1B–$16.8Bfraud, dispute, failure + exception loss
Machine-service revenue: $188M–$674Mnew metered work; may overlap other flows

2032$1.9T
addressable baseline

Agent influenced GMV: $496B–$778Bdiscovery or decision shaped
Agent executed GMV: $192B–$379Bauthority reaches commitment
Merchant-routed value: $170B–$340Bafter modeled channel tolls + loss; not profit
Agent-channel value: $6.1B–$17.1Baccess, orchestration + distribution pool
Risk leakage: $10.0B–$38.3Bfraud, dispute, failure + exception loss
Machine-service revenue: $413M–$1.8Bnew metered work; may overlap other flows

2036$2.4T
addressable baseline

Agent influenced GMV: $758B–$1.4Tdiscovery or decision shaped
Agent executed GMV: $327B–$772Bauthority reaches commitment
Merchant-routed value: $284B–$698Bafter modeled channel tolls + loss; not profit
Agent-channel value: $10.8B–$33.8Baccess, orchestration + distribution pool
Risk leakage: $14.2B–$88.3Bfraud, dispute, failure + exception loss
Machine-service revenue: $907M–$4.7Bnew metered work; may overlap other flows

2040$3.1T
addressable baseline

Agent influenced GMV: $1.0T–$2.2Tdiscovery or decision shaped
Agent executed GMV: $491B–$1.3Tauthority reaches commitment
Merchant-routed value: $413B–$1.1Tafter modeled channel tolls + loss; not profit
Agent-channel value: $15.7B–$55.3Baccess, orchestration + distribution pool
Risk leakage: $22.9B–$168Bfraud, dispute, failure + exception loss
Machine-service revenue: $1.6B–$9.6Bnew metered work; may overlap other flows

Agentic commerce paths · 2029 → 2040

Six futures can grow from the same need.

four stages per path
activation signals can coexist
experiences = illustrative

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2032
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Assistants learn to carry merchant state

Product results become structured enough for an assistant to preserve configuration, availability and intent before handing the person to a merchant.

What purchasing feels like hereThe assistant compares options and opens the exact configured merchant page; the person still reviews and completes the order there.

Conditional direct annual value pool$3.4B$1.0B–$7.9B sensitivity · based on agent-executed GMV at a 2.5% scenario take. Non-additive.

What must become true nextReliable catalog, identity and checkout state can survive the handoff.

How this path can stallMerchants block automated access or expose inconsistent product and fulfillment state.

organic institutionThis path advances through adapters and partial standards; translation expands reach but remains a recurring cost and source of failure.

Range sensitivity0% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationplatform-to-business capability

A different purchasing world · illustrative protocol-enabled purchasing experience

The assistant is the storefront

YouState the outcomeDescribe the care coordination result, budget and non-negotiable constraints.

AIDiscover capabilitiesIntersect the assistant profile with merchant catalog, checkout, fulfillment and identity capabilities.

BizCompose the orderThe merchant returns its own live configuration, price, delivery and support state.

YouApprove in contextOne bounded token authorizes this merchant, amount and order state.

BizFulfill and supportThe merchant stays responsible for the order, delivery, returns and relationship.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:A merchant-owned order is completed through one assistant-rendered experience; fulfillment and support remain with the business.

Mechanism dossier · conditional scenario

Businesses expose typed commerce capabilities while the assistant renders the journey and the merchant keeps the order.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

UCPACPMCP / REST

Activation signal44%

2040 direct value pool$21.5B

Dominant in range draws38%

Human surfaceone composed approval inside the assistant

Control remains hereApprove the composed order and open the merchant’s embedded surface whenever brand, fit or service judgment matters.

What could break it: A business can expose incomplete capabilities or use an embedded experience to reintroduce dark patterns.

All six mechanisms across 512 draws38% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2032
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

One-use authority becomes legible

Agent payments, identity assertions and receipts begin recording why a purchase was allowed, not only that a payment succeeded.

What purchasing feels like hereThe person approves a tightly scoped purchase token and receives a receipt that links intent, seller, amount and result.

Conditional direct annual value pool$952M$204M–$2.9B sensitivity · based on agent-executed GMV at a 0.7% scenario take. Non-additive.

What must become true nextWallets and merchants can represent purpose, scope, expiry and revocation consistently.

How this path can stallMandates remain unreadable legal objects or are reduced to reusable payment credentials.

organic institutionThis path advances through adapters and partial standards; translation expands reach but remains a recurring cost and source of failure.

Range sensitivity74% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationcredentialed agent-to-merchant request

A different purchasing world · illustrative protocol-enabled purchasing experience

Permission replaces checkout

YouGrant a scoped mandateAuthorize this class of care coordination outcome with explicit limits and expiry.

AIPresent minimum credentialsProve identity, eligibility and authority without disclosing the complete personal profile.

BizVerify before reservingValidate the signed intent and bind the proposed order to that authority.

NetSettle with evidencePayment and checkout receipts preserve what was authorized and what executed.

YouReview exceptions onlyRoutine compliant uses stay quiet; boundary crossings return with an explanation.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:A signed chain links the person’s authority to the exact commitment, payment and receipt.

Mechanism dossier · conditional scenario

Signed, scoped authority travels farther than any single agent, so commerce begins with a mandate and ends with auditable evidence.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

AP2Trusted AgentVerifiable Credentials

Activation signal45%

2040 direct value pool$6.0B

Dominant in range draws58%

Human surfaceauthority wallet, exceptions and receipts

Control remains hereEdit or revoke the mandate; approve only uses that cross its amount, identity, timing or consequence boundary.

What could break it: A technically valid mandate can still encode a misunderstood preference or grant more downstream delegation than intended.

All six mechanisms across 512 draws58% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2032
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Adapters translate fragmented commerce

Buyer agents normalize merchant pages and APIs into partial offer schemas, but discovery, logistics and recourse still break at network boundaries.

What purchasing feels like hereThe person receives a cross-site shortlist, then approves a bundle assembled through several visible handoffs.

Conditional direct annual value pool$1.9B$447M–$5.1B sensitivity · based on agent-influenced GMV at a 0.6% scenario take. Non-additive.

What must become true nextBasic offer, fulfillment and receipt semantics converge enough to translate.

How this path can stallEvery large participant extends the schema differently and adapters become permanent tollbooths.

organic institutionThis path advances through adapters and partial standards; translation expands reach but remains a recurring cost and source of failure.

Range sensitivity26% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationfederated demand-and-supply network

A different purchasing world · illustrative protocol-enabled purchasing experience

Commerce becomes a routing network

YouPublish bounded intentSend a time-limited care coordination request without selecting a marketplace first.

NetRoute across networksGateways find compatible seller, logistics, financing and assurance participants.

BizReturn interoperable offersParticipants answer the same schema with inventory, terms, evidence and service capacity.

AICompose the outcomeRank whole outcomes rather than isolated products and preserve independent alternatives.

NetCoordinate fulfillmentCommitments and status updates travel across the selected participant graph.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:The selected bundle can combine inventory, service and physical execution from participants that never shared one marketplace owner.

Mechanism dossier · conditional scenario

Demand is broadcast into interoperable buyer, seller, logistics and assurance networks instead of sent to one seller agent.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

BecknONDCUCP discovery

Activation signal38%

2040 direct value pool$9.3B

Dominant in range draws4%

Human surfacea composed outcome from many network participants

Control remains hereChoose the routing and ranking policy, inspect participant provenance and exclude any network or seller class.

What could break it: Fragmented governance can make responsibility ambiguous when several participants jointly create a failed outcome.

All six mechanisms across 512 draws4% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2040
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Agents buy small pieces of digital work

Machine payments make data, search, verification and generation callable without a subscription or procurement cycle.

What purchasing feels like hereThe person approves a capped task; the assistant buys a few services and shows cost beside the resulting recommendation.

Conditional direct annual value pool$376M$188M–$674M sensitivity · based on metered machine services. Non-additive.

What must become true nextCapabilities can quote, authenticate, settle and return useful evidence cheaply.

How this path can stallMicro-payments add friction without proving that the work improved the outcome.

organic institutionThis path advances through adapters and partial standards; translation expands reach but remains a recurring cost and source of failure.

Range sensitivity0% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationmachine-payable service graph

A different purchasing world · illustrative protocol-enabled purchasing experience

Everything becomes a paid capability

YouFund an outcome reservePlace a capped reserve behind the care coordination policy rather than prepaying one bundle.

AIDiscover paid capabilitiesFind machine-readable services for evidence, planning, capacity and execution.

NetOpen metered sessionsAuthorize capped sessions and stream settlement as useful work is verified.

AIRecompose continuouslyReplace underperforming services without discarding the completed work or policy.

YouAudit spend and resultSee cost by capability, final evidence and every automatically stopped session.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:The reserve is charged as verified work arrives, allowing the service graph to change without rebuilding a cart.

Mechanism dossier · conditional scenario

The commercial unit becomes a session, result or stream of work rather than a cart containing a finished product.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

MPPx402MCP

Activation signal14%

2040 direct value pool$4.0B

Dominant in range draws0%

Human surfacea reserve, live spend stream and result ledger

Control remains hereSet the reserve, per-service ceilings and stop conditions; review the live spend stream or terminate every session.

What could break it: Many individually cheap calls can create runaway spend, dependency chains or incentives to prolong the task.

All six mechanisms across 512 draws0% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2040
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Agents inherit platform reputation

Assistants can use marketplace ratings and escrow, but identity, ranking and dispute history remain trapped inside each platform.

What purchasing feels like hereThe person sees which platform guarantee backs the recommendation and must enter that platform to resolve failure.

Conditional direct annual value pool$1.6B$306M–$4.8B sensitivity · based on agent-executed GMV at a 1.2% scenario take. Non-additive.

What must become true nextAgent actions can be attributed to verified accounts and bounded payment instruments.

How this path can stallSynthetic reviews and account farming make inherited reputation less useful to machines.

organic institutionThis path advances through adapters and partial standards; translation expands reach but remains a recurring cost and source of failure.

Range sensitivity0% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationregistry-and-escrow work market

A different purchasing world · illustrative protocol-enabled purchasing experience

Trust becomes the marketplace

AIPost a verifiable jobPublish the care coordination result, acceptance tests, deadline and escrow terms.

NetFilter identity and reputationResolve registered providers, attestations, prior outcomes and required bonds.

BizBid with evidenceProviders commit price, method, service terms and a bond against non-performance.

NetEvaluate the outcomeIndependent evaluators check delivery evidence and open a dispute window.

YouAccept or appealEscrow releases automatically inside policy; contested judgments return with the full record.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:Unknown providers can compete without requiring one platform to own identity, ranking, payment and dispute resolution.

Mechanism dossier · conditional scenario

Identity, reputation, escrow and independent evaluation let unknown agents form temporary markets around a job.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

ERC-8004ERC-8183A2A

Activation signal35%

2040 direct value pool$10.3B

Dominant in range draws0%

Human surfaceproof thresholds, bonds, evaluators and recourse

Control remains hereChoose reputation sources, minimum bonds, evaluators and appeal rules before the job is posted.

What could break it: Reputation can be gamed, evaluators can collude and on-chain evidence can preserve a wrong judgment permanently.

All six mechanisms across 512 draws0% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2040
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Products expose a persistent identity

Digital links and event histories let a person or assistant retrieve authoritative product, ownership and service information after purchase.

What purchasing feels like hereThe person scans or asks the assistant about the asset; service history and compatible options replace a fresh web search.

Conditional direct annual value pool$144M$35.8M–$358M sensitivity · based on physically orchestrated GMV at a 3.0% scenario take. Non-additive.

What must become true nextProduct identifiers resolve to portable, owner-accessible lifecycle data.

How this path can stallManufacturers control the history, withhold repair data or abandon the resolver.

organic institutionThis path advances through adapters and partial standards; translation expands reach but remains a recurring cost and source of failure.

Range sensitivity0% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationasset-to-service coordination loop

A different purchasing world · illustrative protocol-enabled purchasing experience

Physical objects become the buyers

ObjDetect a state changeThe asset records the care coordination condition with time, location, provenance and sensor evidence.

ObjCheck its service policyDetermine whether to observe, reserve capacity, request approval or stop operation.

NetFind compatible physical serviceQuery parts, technicians, facilities and logistics against the asset identity and history.

AIAuthorize the responseApply the owner’s mandate, preserve alternatives and bind the selected physical plan.

ObjVerify the new stateRecord completion evidence, update the twin and release payment only for the verified result.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:The asset’s condition, service history and verified result form one continuing commercial record instead of separate shopping and maintenance journeys.

Mechanism dossier · conditional scenario

Assets publish their own condition and service history, allowing a vehicle, appliance, building or inventory item to initiate commerce.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

GS1 Digital LinkEPCISdigital twins

Activation signal20%

2040 direct value pool$15.2B

Dominant in range draws0%

Human surfacea living asset timeline with revocable operating policy

Control remains hereControl which sensors and service classes may initiate action, require approval for irreversible work and revoke the asset mandate.

What could break it: Bad sensor data, compromised twins or manufacturer-controlled histories can trigger unnecessary work or lock owners into captive service.

All six mechanisms across 512 draws0% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From one assistant to an agent ecology

Eight forms can coexist—and diverge.

strongest form outlined
physical implications authored

interaction form2029203220362040

Site referralPhysical fulfillment remains downstream of a conventional order.

79%

64%

49%

39%

Delegated checkoutThe seller fulfills an agent-authored order rather than a person-built cart.

17%

26%

32%

34%

Bilateral agent tradeInventory, service capacity and delivery can be reserved machine-to-machine.

10%

16%

Terms marketThe physical plan changes before checkout because timing and capacity are part of the bargain.

10%

Standing outcome policySupply, repair, transport or care coordination continue without a new shopping session.

Collective procurementNeighborhoods, fleets or firms reshape routes, inventory and capacity through coordinated demand.

11%

Commissioned supplyProduction schedules, components and facilities respond to intent rather than forecasted shelf demand.

10%

Embodied orchestrationThe purchasing interface becomes a live map of actions, safety boundaries and revocable authority.

16%

Delegation frontier · published path

Where autonomy stops depends on the need.

engine-selected interface + full range coverage · unanchored priors · illustrative UI

commerce arena2029203220362040

Household replenishmentexplore this needapproval100% selectednegotiate73% selectedpolicy69% selectedambient67% selected

Everyday mobilityexplore this needapproval100% selectedapproval75% selectednegotiate69% selectedpolicy66% selected

Complex travelexplore this needapproval100% selectedapproval100% selectedapproval76% selectednegotiate66% selected

Care coordinationselected lensapproval100% selectedapproval100% selectedapproval76% selectednegotiate66% selected

approvalnegotiatepolicyambient

Published path · structural model

Walled gardens

Platform tokens are exclusive. Chat runtimes block third-party rails almost half the time, take rates run at a tenth or more, and the open protocols cost a merchant three times as much to integrate as the platform's own checkout. Agents still grow — 15% of online GMV by 2029Q2 — but on cards and platform checkout, with wallets stuck below a fifth of buyers and cross-garden purchases a coin flip.

2040 machine counterparties16%

continuous commerce4%

published 2040 p10–p908%–24%512 structural draws

The whole path in four changes

One need moves from request to operating policy.

story first · model layers after

2029 · assisted commerceIntent becomes a mandate

The person gives software a bounded right to act.

purchase: approval bundle
agent-executed: $76.4B–$118B
machine cast: 3 role classes
human: operator

2032 · assisted commerceBusinesses become callable

Catalog, checkout, fulfillment and support become machine-addressable capabilities.

purchase: approval bundle
agent-executed: $150B–$310B
machine cast: 4 role classes
human: operator

2036 · delegated commerceThe market gains counterparties

Specialized buyer, seller, trust and operating agents coordinate the purchase.

purchase: approval bundle
agent-executed: $259B–$639B
machine cast: 8 role classes
human: approver and recourse owner

2040 · delegated commerceThe purchase becomes a policy

Standing intent allocates money and physical capacity until a boundary is crossed.

purchase: approval bundle
agent-executed: $397B–$1.1T
machine cast: 9 role classes
human: approver and recourse owner

Conditional annual ranges overlap and are not valuations. Role classes are functional activations above the disclosed threshold—not agent population or forecast headcount.

2029

assisted commerce · model-triggered authored event

Distribution turns checkout into a platform boundary.

Integrated runtimes make agent checkout coherent inside their own marketplaces while third-party reach remains conditional.

2029

Illustrative Care coordination purchase · Walled gardens

One intent, one approval

State the goal, inspect the evidence and approve this purchase.

approval bundle · engine selected

Agent influenced$234B–$268B

Agent executed$76.4B–$118B

Merchant-routed after tolls + loss$62.7B–$98.1B

CCAuthorization bundleCare coordination · authenticated workspaceready for approvalWalled gardens2029

IntentFind three eligible care providers with the right specialty, availability and language fit; do not book without approval.

Recommended bundleEligible providers · language fit · verified availabilityCoverage, credentials and continuity evidence remain inspectable · runtime-bound account · platform guarantee

within mandate

FallbackEarlier intake · different locationNever booked without explicit provider approval · invoke the platform guarantee or appeal inside the runtime, with no portable fallback if the account is suspended

person decides

One signed commitment · one-shot settlementAuthorize purchase

terms verifiedmandate · CARE-2029-04proof attached

Illustrative outcomeThe person approves one bounded commitment and receives a receipt plus execution watch.

intentsourcenegotiatesettleassure

personsharedagent

Authoritysingle-use mandate

Market topologymanaged marketplace

Identityruntime-bound account

Price formationposted bundle

Settlementone-shot settlement

Human attentionapprove this purchase

Trust surfaceruntime-bound account, platform guarantee, provenance and a receipt that records the mandate and result.

How the market respondsAssembles a ranked offer, counterparty, settlement and assurance bundle, then waits for one explicit authorization. Sourcing uses ranked shortlist from a managed marketplace; posted bundles set the price before one-shot settlement executes the commitment.

When it breaksinvoke the platform guarantee or appeal inside the runtime, with no portable fallback if the account is suspended; the person’s attention mode is “approve this purchase.”

autonomy5%

delegation fit13%

market reach46%

exception risk38%

Range-covering interface spread

approval 100%negotiate 0%policy 0%ambient 0%

512 structural draws · sensitivity coverage, not probability

Evidence labelIndicators combine structural state with disclosed, unanchored arena priors. Dialogue and interface remain illustrative, not forecast outputs.

In the physical world · illustrative physical-world experience

The agent prepares the visit; the person chooses the provider.

2029 · Walled gardens
managed marketplace
Site referral 84%

What the person experiencesThe visible moment

The person sees three eligible providers with verified specialty, language fit, availability and coverage evidence.

What software does off-screenThe delegated work

The agent verifies administrative facts, resolves contradictory directories and prepares—but does not submit—the booking.

What changes in the worldThe physical handoff

After explicit approval, the clinic receives the minimum eligibility and scheduling information needed for intake.

Control returns hereProvider selection, booking and every clinical choice remain with a person.

What happens next · model-derived interpretation

The visible change is compression: product search, seller choice, service, payment and delivery arrive as one proposal. Human attention concentrates at the moment of authorization. At the 2029 handoff, 15% of demand is delegated while only 1% has a machine on both sides. The experience therefore begins as an approval bundle.

What remains contested: The managed marketplace makes routine execution coherent, but a runtime-bound account concentrates ranking, appeal and exit inside the runtime.

Active machine cast · model-derivedBuyer agent · Mandate agent · Settlement agent

Coordination burden3%range 2%–3%

Human roleoperatorowns multi-agent drift

Between purchasing chapters · model-derived experience transition

The structural handoff begins as an approval bundle.

2029Q2 → 2029

Interfaceapproval bundlestructural starting state

Human attentionapprove this purchasestructural starting state

Agent mandatesingle-use mandateautonomy starts at 5%

Market boundary46% reachablereach starts at 46% · exception risk starts at 38%

Model trace · why this chapter changes

Model trigger · branch-specific event at the near-term handoff15.5% of demand is delegated, while seller-agent adoption begins at 7.7% after the 2029Q2 handoff.Authored consequence: Merchant access to delegated demand now depends on platform distribution and ranking policy.

Event sensitivity · branch-fixed handoff event

Distribution turns checkout into a platform boundary.100%

512 range-covering structural draws · authored event selection coverage, not probability. Distribution turns checkout into a platform boundary. 100%

delegated demand15%

machine on both sides1%

continuous commerce2%

physical execution0%

pooled demand0%

adaptive supply0%

governance capacity37%

adversarial loss9%

2029Q2 → structural 2029model-derived transition · equations, not observed attribution

delegated demandThe structural layer carries the branch’s final agent-GMV share into its starting delegated-demand state.15%near-term agent GMV 15% → 15%

seller-agent footholdA disclosed 12% handoff factor converts rail-capable sellers into the initial seller-agent foothold.8%sellers answering a rail 65% → 8%

governance capacityPortable authority, liability clarity and branch coordination are averaged into the initial governance capacity.37%structural priors

physical execution footholdA disclosed handoff factor seeds the small share of agent demand that can already cause and verify physical execution.0%near-term agent GMV + structural prior 15% → 0%

pooled demand footholdDelegated demand and the pooled-mandate prior seed the initial ability to bargain across more than one buyer.0%agent GMV + pooled-mandate prior 15% → 0%

adaptive supply footholdA disclosed handoff factor seeds supply that can be configured from machine-readable demand rather than selected from a fixed catalog.0%seller-agent foothold

adversarial lossThe failed-attempt share and adversarial-pressure prior seed loss pressure; this is a structural mapping, not observed loss.9%failed attempts + pressure prior 83% → 9%

share of demand delegated starts at 15.5%; seller-agent adoption starts at 7.7%; share with machine counterparties starts at 1.2%; share negotiated starts at 1.2%; share run continuously starts at 1.5%.

No new actor class enters yet; adoption and institutional state drive this transition.

modeled from unanchored structural inputs · 17 published research hooks · ensemble ranges are modeled sensitivity, not measured probability

2032

assisted commerce · model-triggered authored event

Authority remains attached to the runtime account.

Runtime accounts become the default container for spending authority, commercial history and the platform guarantee.

2032

Illustrative Care coordination purchase · Walled gardens

One intent, one approval

State the goal, inspect the evidence and approve this purchase.

approval bundle · engine selected

Agent influenced$412B–$670B

Agent executed$150B–$310B

Merchant-routed after tolls + loss$126B–$262B

CCAuthorization bundleCare coordination · authenticated workspaceready for approvalWalled gardens2032

IntentNegotiate intake timing and verified coverage while preserving provider choice and continuity constraints.

Recommended bundleEligible providers · language fit · verified availabilityCoverage, credentials and continuity evidence remain inspectable · runtime-bound account · platform guarantee

within mandate

person decides

One signed commitment · one-shot settlementAuthorize purchase

terms verifiedmandate · CARE-2032-04proof attached

Illustrative outcomeThe person approves one bounded commitment and receives a receipt plus execution watch.

intentsourcenegotiatesettleassure

personsharedagent

Authoritysingle-use mandate

Market topologymanaged marketplace

Identityruntime-bound account

Price formationposted bundle

Settlementone-shot settlement

Human attentionapprove this purchase

Trust surfaceruntime-bound account, platform guarantee, provenance and a receipt that records the mandate and result.

How the market respondsAssembles a ranked offer, counterparty, settlement and assurance bundle, then waits for one explicit authorization. Sourcing uses ranked shortlist from a managed marketplace; posted bundles set the price before one-shot settlement executes the commitment.

When it breaksinvoke the platform guarantee or appeal inside the runtime, with no portable fallback if the account is suspended; the person’s attention mode is “approve this purchase.”

autonomy8%

delegation fit13%

market reach46%

exception risk38%

Range-covering interface spread

approval 100%negotiate 0%policy 0%ambient 0%

512 structural draws · sensitivity coverage, not probability

Evidence labelIndicators combine structural state with disclosed, unanchored arena priors. Dialogue and interface remain illustrative, not forecast outputs.

In the physical world · illustrative physical-world experience

Administrative terms negotiate; care decisions do not.

2032 · Walled gardens
managed marketplace
Site referral 72%

What the person experiencesThe visible moment

The person sees an earlier intake at a new location beside a later appointment that preserves continuity and language fit.

What software does off-screenThe delegated work

Provider, benefits and transport agents negotiate eligibility, intake timing and logistics without exchanging clinical discretion.

What changes in the worldThe physical handoff

The chosen clinic, ride and benefits authorization are coordinated only after the person selects the provider path.

Control returns hereThe mandate excludes diagnosis, treatment and autonomous provider substitution.

What happens next · model-derived interpretation

The visible change is compression: product search, seller choice, service, payment and delivery arrive as one proposal. Human attention concentrates at the moment of authorization. Since 2029, machine-counterparty share rises by 3.4 percentage points and continuous commerce falls by 0.3 percentage points. The experience remains an approval bundle, but its mandate and market scope deepen.

What remains contested: The managed marketplace makes routine execution coherent, but a runtime-bound account concentrates ranking, appeal and exit inside the runtime.

Active machine cast · model-derivedSeller agent · Buyer agent · Mandate agent · Settlement agent

Coordination burden4%range 2%–4%

Human roleoperatorowns multi-agent drift

Between purchasing chapters · model-derived experience transition

approval bundle remains visible while its mandate deepens.

2029 → 2032

Interfaceapproval bundleunchanged from approval bundle

Human attentionapprove this purchaseunchanged from approve this purchase

Agent mandatesingle-use mandateautonomy +3.2 pp to 8%

Market boundary46% reachablereach -0.0 pp to 46% · exception risk -0.0 pp to 38%

Model trace · why this chapter changes

Model trigger · identity portability < 35.0%Identity portability is 18.0% and machine counterparties reach 4.6%.Authored consequence: The guarantee reduces transaction friction while making suspension, appeal and exit economically consequential.

Event sensitivity · selected 31% · leading alternate 38%

Authority remains attached to the runtime account.31%

Platforms license selective credential bridges.38%

Regulation forces portable mandates out of the garden.31%

512 range-covering structural draws · authored event selection coverage, not probability. Authority remains attached to the runtime account. 31% · Platforms license selective credential bridges. 38% · Regulation forces portable mandates out of the garden. 31%

delegated demand25%

machine on both sides5%

continuous commerce1%

physical execution5%

pooled demand1%

adaptive supply3%

governance capacity37%

adversarial loss9%

2029 → 2032model-derived transition · equations, not observed attribution

delegated demandCommercial capability closes part of the gap toward the governance- and loss-constrained trust ceiling.+9.3 pp2029 15% → now 25%

seller-agent adoptionSeller-side diffusion follows the demand agents can address and the selected annual diffusion prior.+25.8 pp2029 8% → now 34%

machine counterpartiesBuyer and seller delegation meet at a rate shaped by persistent-agent cost and branch coordination.+3.4 pp2029 1% → now 5%

continuous commerceDelegated demand, machine counterparties and governance turn isolated transactions into standing policies.−0.3 pp2029 2% → now 1%

physical executionAgent-addressable operators diffuse toward a ceiling set by seller agents, machine counterparties and governance capacity.+5.2 pp2029 0% → now 5%

pooled demandPortable pooled mandates, branch coordination and market concentration determine how much delegated demand can bargain collectively.+1.1 pp2029 0% → now 1%

adaptive supplySeller agents, negotiated transactions and physical execution let supply respond to intent before a finished item exists.+2.8 pp2029 0% → now 3%

adversarial lossFraud and manipulation pressure compete with governance capacity as machine counterparties expand.−0.0 pp2029 9% → now 9%

share of demand delegated rises to 24.8% (+9.3%); seller-agent adoption rises to 33.5% (+25.8%); share with machine counterparties rises to 4.6% (+3.4%); share negotiated rises to 2.1% (+0.9%); share run continuously falls to 1.2% (-0.3%).

modeled from unanchored structural inputs · 17 published research hooks · ensemble ranges are modeled sensitivity, not measured probability

2036

delegated commerce · model-triggered authored event

Platform guarantees keep approval in the loop.

Buyer and seller agents negotiate inside platform-run markets whose operator controls access, ranking and acceptable terms.

2036

Illustrative Care coordination purchase · Walled gardens

One intent, one approval

State the goal, inspect the evidence and approve this purchase.

approval bundle · engine selected

Agent influenced$656B–$1.3T

Agent executed$259B–$639B

Merchant-routed after tolls + loss$213B–$555B

CCAuthorization bundleCare coordination · authenticated workspaceready for approvalWalled gardens2036

IntentMaintain care continuity below the benefit ceiling; coordinate eligibility, appointments and transport, escalating every clinical choice.

Recommended bundleEligible providers · language fit · verified availabilityCoverage, credentials and continuity evidence remain inspectable · runtime-bound account · platform guarantee

within mandate

person decides

One signed commitment · conditional atomic settlementAuthorize purchase

terms verifiedmandate · CARE-2036-04proof attached

Illustrative outcomeThe person approves one bounded commitment and receives a receipt plus execution watch.

intentsourcenegotiatesettleassure

personsharedagent

Authoritysingle-use mandate

Market topologymanaged marketplace

Identityruntime-bound account

Price formationnegotiated bundle

Settlementconditional atomic settlement

Human attentionapprove this purchase

Trust surfaceruntime-bound account, platform guarantee, provenance and a receipt that records the mandate and result.

How the market respondsAssembles a ranked offer, counterparty, settlement and assurance bundle, then waits for one explicit authorization. Sourcing uses ranked shortlist from a managed marketplace; buyer and seller agents negotiate the bundle before conditional atomic settlement executes the commitment.

When it breaksinvoke the platform guarantee or appeal inside the runtime, with no portable fallback if the account is suspended; the person’s attention mode is “approve this purchase.”

autonomy12%

delegation fit13%

market reach46%

exception risk39%

Range-covering interface spread

approval 96%negotiate 4%policy 0%ambient 0%

512 structural draws · sensitivity coverage, not probability

Surface sensitivityThe engine-selected surface appears in 96% of the range-covering draws above. The remaining share shows which alternate purchasing interfaces the same model selects as structural assumptions move.

Evidence labelIndicators combine structural state with disclosed, unanchored arena priors. Dialogue and interface remain illustrative, not forecast outputs.

In the physical world · illustrative physical-world experience

Care logistics become continuous while clinical authority stays human.

2036 · Walled gardens
managed marketplace
Site referral 59%

What the person experiencesThe visible moment

A single care timeline shows appointments, eligibility, transport and benefit headroom, with a provider-change exception waiting.

What software does off-screenThe delegated work

The agent reconciles administrative commitments across clinics, benefits and mobility, and pauses whenever a clinical boundary appears.

What changes in the worldThe physical handoff

Transport, intake paperwork and coverage confirmations arrive in sequence around the person’s chosen care plan.

Control returns hereThe person and clinicians retain every diagnosis, treatment and provider decision; the agent coordinates only the surrounding work.

What happens next · model-derived interpretation

The visible change is compression: product search, seller choice, service, payment and delivery arrive as one proposal. Human attention concentrates at the moment of authorization. Since 2032, machine-counterparty share rises by 6.3 percentage points and continuous commerce rises by 0.9 percentage points. The experience remains an approval bundle, but its mandate and market scope deepen.

What remains contested: The managed marketplace makes routine execution coherent, but a runtime-bound account concentrates ranking, appeal and exit inside the runtime.

Active machine cast · model-derivedSeller agent · Buyer agent · Mandate agent · Settlement agent · Physical operator · Routing agent · Market operator · Supply agent

Coordination burden6%range 3%–8%

Human roleapprover and recourse ownerowns multi-agent drift

Between purchasing chapters · model-derived experience transition

approval bundle remains visible while its mandate deepens.

2032 → 2036

Interfaceapproval bundleunchanged from approval bundle

Human attentionapprove this purchaseunchanged from approve this purchase

Agent mandatesingle-use mandateautonomy +4.1 pp to 12%

Market boundary46% reachablereach -0.1 pp to 46% · exception risk +0.9 pp to 39%

Model trace · why this chapter changes

Model trigger · negotiated transactions < 6.0%4.8% of transactions are negotiated and seller-agent adoption reaches 53.0%.Authored consequence: Negotiation becomes efficient inside the garden, but the operator can observe and shape both sides of price formation.

Event sensitivity · selected 80% · leading alternate 20%

Platform guarantees keep approval in the loop.80%

Runtime agents bargain inside captive markets.20%

Platform auctions internalize machine negotiation.0%

512 range-covering structural draws · authored event selection coverage, not probability. Platform guarantees keep approval in the loop. 80% · Runtime agents bargain inside captive markets. 20% · Platform auctions internalize machine negotiation. 0%

delegated demand34%

machine on both sides11%

continuous commerce2%

physical execution15%

pooled demand2%

adaptive supply9%

governance capacity37%

adversarial loss10%

2032 → 2036model-derived transition · equations, not observed attribution

delegated demandCommercial capability closes part of the gap toward the governance- and loss-constrained trust ceiling.+8.8 pp2032 25% → now 34%

seller-agent adoptionSeller-side diffusion follows the demand agents can address and the selected annual diffusion prior.+19.4 pp2032 34% → now 53%

machine counterpartiesBuyer and seller delegation meet at a rate shaped by persistent-agent cost and branch coordination.+6.3 pp2032 5% → now 11%

continuous commerceDelegated demand, machine counterparties and governance turn isolated transactions into standing policies.+0.9 pp2032 1% → now 2%

physical executionAgent-addressable operators diffuse toward a ceiling set by seller agents, machine counterparties and governance capacity.+9.8 pp2032 5% → now 15%

pooled demandPortable pooled mandates, branch coordination and market concentration determine how much delegated demand can bargain collectively.+0.9 pp2032 1% → now 2%

adaptive supplySeller agents, negotiated transactions and physical execution let supply respond to intent before a finished item exists.+5.9 pp2032 3% → now 9%

adversarial lossFraud and manipulation pressure compete with governance capacity as machine counterparties expand.+0.9 pp2032 9% → now 10%

share of demand delegated rises to 33.6% (+8.8%); seller-agent adoption rises to 53.0% (+19.4%); share with machine counterparties rises to 10.9% (+6.3%); share negotiated rises to 4.8% (+2.7%); share run continuously rises to 2.1% (+0.9%).

modeled from unanchored structural inputs · 17 published research hooks · ensemble ranges are modeled sensitivity, not measured probability

2040

delegated commerce · model-triggered authored event

Persistent agents remain bounded by platform guarantees.

Persistent runtime agents manage recurring needs through captive accounts, guarantees and preferred seller capacity.

2040

Illustrative Care coordination purchase · Walled gardens

One intent, one approval

State the goal, inspect the evidence and approve this purchase.

approval bundle · engine selected

Agent influenced$905B–$2.0T

Agent executed$397B–$1.1T

Merchant-routed after tolls + loss$313B–$912B

CCAuthorization bundleCare coordination · authenticated workspaceready for approvalWalled gardens2040

IntentKeep the care plan coordinated across providers and benefits while reserving all diagnosis, treatment and provider changes for a person.

Recommended bundleEligible providers · language fit · verified availabilityCoverage, credentials and continuity evidence remain inspectable · runtime-bound account · platform guarantee

within mandate

person decides

One signed commitment · conditional atomic settlementAuthorize purchase

terms verifiedmandate · CARE-2040-04proof attached

Illustrative outcomeThe person approves one bounded commitment and receives a receipt plus execution watch.

intentsourcenegotiatesettleassure

personsharedagent

Authoritysingle-use mandate

Market topologymanaged marketplace

Identityruntime-bound account

Price formationnegotiated bundle

Settlementconditional atomic settlement

Human attentionapprove this purchase

Trust surfaceruntime-bound account, platform guarantee, provenance and a receipt that records the mandate and result.

How the market respondsAssembles a ranked offer, counterparty, settlement and assurance bundle, then waits for one explicit authorization. Sourcing uses ranked shortlist from a managed marketplace; buyer and seller agents negotiate the bundle before conditional atomic settlement executes the commitment.

When it breaksinvoke the platform guarantee or appeal inside the runtime, with no portable fallback if the account is suspended; the person’s attention mode is “approve this purchase.”

autonomy15%

delegation fit13%

market reach45%

exception risk40%

Range-covering interface spread

approval 69%negotiate 31%policy 0%ambient 0%

512 structural draws · sensitivity coverage, not probability

Surface sensitivityThe engine-selected surface appears in 69% of the range-covering draws above. The remaining share shows which alternate purchasing interfaces the same model selects as structural assumptions move.

Evidence labelIndicators combine structural state with disclosed, unanchored arena priors. Dialogue and interface remain illustrative, not forecast outputs.

In the physical world · illustrative physical-world experience

The most capable agent is visible in the moments it refuses to act.

2040 · Walled gardens
managed marketplace
Site referral 50%

What the person experiencesThe visible moment

The person sees six weeks of coordinated access and a clear pause: the next step would change provider continuity.

What software does off-screenThe delegated work

The agent keeps benefits, appointments and transport synchronized, but marks the unresolved decision instead of optimizing through it.

What changes in the worldThe physical handoff

Existing care continues, provisional capacity is held and no new provider receives authority until a person decides.

Control returns hereContinuous administrative agency never becomes continuous clinical agency.

What happens next · model-derived interpretation

The visible change is compression: product search, seller choice, service, payment and delivery arrive as one proposal. Human attention concentrates at the moment of authorization. Since 2036, machine-counterparty share rises by 5.5 percentage points and continuous commerce rises by 1.6 percentage points. The experience remains an approval bundle, but its mandate and market scope deepen.

What remains contested: The managed marketplace makes routine execution coherent, but a runtime-bound account concentrates ranking, appeal and exit inside the runtime.

Active machine cast · model-derivedSeller agent · Buyer agent · Mandate agent · Settlement agent · Physical operator · Routing agent · Market operator · Supply agent · Assurance agent

Coordination burden9%range 5%–10%

Human roleapprover and recourse ownerowns coordination capture

Between purchasing chapters · model-derived experience transition

approval bundle remains visible while its mandate deepens.

2036 → 2040

Interfaceapproval bundleunchanged from approval bundle

Human attentionapprove this purchaseunchanged from approve this purchase

Agent mandatesingle-use mandateautonomy +3.3 pp to 15%

Market boundary45% reachablereach -0.1 pp to 45% · exception risk +1.1 pp to 40%

Model trace · why this chapter changes

Model trigger · continuous commerce < 12.0%3.6% of commerce runs continuously while market concentration is 78.8%.Authored consequence: Execution feels continuous to the buyer while switching costs accumulate in identity, history, reserves and redress.

Event sensitivity · selected 99% · leading alternate 1%

Persistent agents remain bounded by platform guarantees.99%

Runtime subscriptions absorb the checkout session.1%

Platform reserves become private outcome markets.0%

512 range-covering structural draws · authored event selection coverage, not probability. Persistent agents remain bounded by platform guarantees. 99% · Runtime subscriptions absorb the checkout session. 1% · Platform reserves become private outcome markets. 0%

delegated demand40%

machine on both sides16%

continuous commerce4%

physical execution23%

pooled demand3%

adaptive supply14%

governance capacity37%

adversarial loss11%

2036 → 2040model-derived transition · equations, not observed attribution

delegated demandCommercial capability closes part of the gap toward the governance- and loss-constrained trust ceiling.+6.0 pp2036 34% → now 40%

seller-agent adoptionSeller-side diffusion follows the demand agents can address and the selected annual diffusion prior.+7.7 pp2036 53% → now 61%

machine counterpartiesBuyer and seller delegation meet at a rate shaped by persistent-agent cost and branch coordination.+5.5 pp2036 11% → now 16%

continuous commerceDelegated demand, machine counterparties and governance turn isolated transactions into standing policies.+1.6 pp2036 2% → now 4%

physical executionAgent-addressable operators diffuse toward a ceiling set by seller agents, machine counterparties and governance capacity.+7.5 pp2036 15% → now 23%

pooled demandPortable pooled mandates, branch coordination and market concentration determine how much delegated demand can bargain collectively.+0.7 pp2036 2% → now 3%

adaptive supplySeller agents, negotiated transactions and physical execution let supply respond to intent before a finished item exists.+5.1 pp2036 9% → now 14%

adversarial lossFraud and manipulation pressure compete with governance capacity as machine counterparties expand.+1.1 pp2036 10% → now 11%

share of demand delegated rises to 39.7% (+6.0%); seller-agent adoption rises to 60.6% (+7.7%); share with machine counterparties rises to 16.4% (+5.5%); share negotiated rises to 7.9% (+3.1%); share run continuously rises to 3.6% (+1.6%).

modeled from unanchored structural inputs · 17 published research hooks · ensemble ranges are modeled sensitivity, not measured probability

Expand the forces behind this futureeconomic envelope · six protocol paths · full agent ecology

Economic envelope · annual U.S. retail e-commerce flows

A dollar can be influenced, executed, retained—or captured as infrastructure.

512 paired range-covering draws
one observed 2026 anchor
not a valuation or probability

agent influencedagent executedmerchant-routed after channel value + loss

2029$1.6T
addressable baseline

Agent influenced GMV: $234B–$268Bdiscovery or decision shaped
Agent executed GMV: $76.4B–$118Bauthority reaches commitment
Merchant-routed value: $62.7B–$98.1Bafter modeled channel tolls + loss; not profit
Agent-channel value: $4.4B–$11.0Baccess, orchestration + distribution pool
Risk leakage: $4.8B–$13.1Bfraud, dispute, failure + exception loss
Machine-service revenue: $137M–$493Mnew metered work; may overlap other flows

2032$1.9T
addressable baseline

Agent influenced GMV: $412B–$670Bdiscovery or decision shaped
Agent executed GMV: $150B–$310Bauthority reaches commitment
Merchant-routed value: $126B–$262Bafter modeled channel tolls + loss; not profit
Agent-channel value: $9.5B–$27.2Baccess, orchestration + distribution pool
Risk leakage: $8.5B–$31.5Bfraud, dispute, failure + exception loss
Machine-service revenue: $311M–$1.4Bnew metered work; may overlap other flows

2036$2.4T
addressable baseline

Agent influenced GMV: $656B–$1.3Tdiscovery or decision shaped
Agent executed GMV: $259B–$639Bauthority reaches commitment
Merchant-routed value: $213B–$555Bafter modeled channel tolls + loss; not profit
Agent-channel value: $17.1B–$55.0Baccess, orchestration + distribution pool
Risk leakage: $12.5B–$70.4Bfraud, dispute, failure + exception loss
Machine-service revenue: $677M–$3.5Bnew metered work; may overlap other flows

2040$3.1T
addressable baseline

Agent influenced GMV: $905B–$2.0Tdiscovery or decision shaped
Agent executed GMV: $397B–$1.1Tauthority reaches commitment
Merchant-routed value: $313B–$912Bafter modeled channel tolls + loss; not profit
Agent-channel value: $25.1B–$90.1Baccess, orchestration + distribution pool
Risk leakage: $19.7B–$128Bfraud, dispute, failure + exception loss
Machine-service revenue: $1.2B–$6.9Bnew metered work; may overlap other flows

Agentic commerce paths · 2029 → 2040

Six futures can grow from the same need.

four stages per path
activation signals can coexist
experiences = illustrative

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2032
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Assistants learn to carry merchant state

Product results become structured enough for an assistant to preserve configuration, availability and intent before handing the person to a merchant.

What purchasing feels like hereThe assistant compares options and opens the exact configured merchant page; the person still reviews and completes the order there.

Conditional direct annual value pool$2.5B$764M–$5.9B sensitivity · based on agent-executed GMV at a 2.5% scenario take. Non-additive.

What must become true nextReliable catalog, identity and checkout state can survive the handoff.

How this path can stallMerchants block automated access or expose inconsistent product and fulfillment state.

walled institutionThis path scales fastest inside integrated platforms; the experience is coherent, but identity, evidence and exit remain captive to the runtime.

Range sensitivity30% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationplatform-to-business capability

A different purchasing world · illustrative protocol-enabled purchasing experience

The assistant is the storefront

YouState the outcomeDescribe the care coordination result, budget and non-negotiable constraints.

AIDiscover capabilitiesIntersect the assistant profile with merchant catalog, checkout, fulfillment and identity capabilities.

BizCompose the orderThe merchant returns its own live configuration, price, delivery and support state.

YouApprove in contextOne bounded token authorizes this merchant, amount and order state.

BizFulfill and supportThe merchant stays responsible for the order, delivery, returns and relationship.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:A merchant-owned order is completed through one assistant-rendered experience; fulfillment and support remain with the business.

Mechanism dossier · conditional scenario

Businesses expose typed commerce capabilities while the assistant renders the journey and the merchant keeps the order.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

UCPACPMCP / REST

Activation signal46%

2040 direct value pool$16.0B

Dominant in range draws71%

Human surfaceone composed approval inside the assistant

Control remains hereApprove the composed order and open the merchant’s embedded surface whenever brand, fit or service judgment matters.

What could break it: A business can expose incomplete capabilities or use an embedded experience to reintroduce dark patterns.

All six mechanisms across 512 draws71% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2032
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

One-use authority becomes legible

Agent payments, identity assertions and receipts begin recording why a purchase was allowed, not only that a payment succeeded.

What purchasing feels like hereThe person approves a tightly scoped purchase token and receives a receipt that links intent, seller, amount and result.

Conditional direct annual value pool$700M$153M–$2.1B sensitivity · based on agent-executed GMV at a 0.7% scenario take. Non-additive.

What must become true nextWallets and merchants can represent purpose, scope, expiry and revocation consistently.

How this path can stallMandates remain unreadable legal objects or are reduced to reusable payment credentials.

walled institutionThis path scales fastest inside integrated platforms; the experience is coherent, but identity, evidence and exit remain captive to the runtime.

Range sensitivity70% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationcredentialed agent-to-merchant request

A different purchasing world · illustrative protocol-enabled purchasing experience

Permission replaces checkout

YouGrant a scoped mandateAuthorize this class of care coordination outcome with explicit limits and expiry.

AIPresent minimum credentialsProve identity, eligibility and authority without disclosing the complete personal profile.

BizVerify before reservingValidate the signed intent and bind the proposed order to that authority.

NetSettle with evidencePayment and checkout receipts preserve what was authorized and what executed.

YouReview exceptions onlyRoutine compliant uses stay quiet; boundary crossings return with an explanation.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:A signed chain links the person’s authority to the exact commitment, payment and receipt.

Mechanism dossier · conditional scenario

Signed, scoped authority travels farther than any single agent, so commerce begins with a mandate and ends with auditable evidence.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

AP2Trusted AgentVerifiable Credentials

Activation signal36%

2040 direct value pool$4.5B

Dominant in range draws29%

Human surfaceauthority wallet, exceptions and receipts

Control remains hereEdit or revoke the mandate; approve only uses that cross its amount, identity, timing or consequence boundary.

What could break it: A technically valid mandate can still encode a misunderstood preference or grant more downstream delegation than intended.

All six mechanisms across 512 draws29% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2032
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Adapters translate fragmented commerce

Buyer agents normalize merchant pages and APIs into partial offer schemas, but discovery, logistics and recourse still break at network boundaries.

What purchasing feels like hereThe person receives a cross-site shortlist, then approves a bundle assembled through several visible handoffs.

Conditional direct annual value pool$1.5B$351M–$4.0B sensitivity · based on agent-influenced GMV at a 0.6% scenario take. Non-additive.

What must become true nextBasic offer, fulfillment and receipt semantics converge enough to translate.

How this path can stallEvery large participant extends the schema differently and adapters become permanent tollbooths.

walled institutionThis path scales fastest inside integrated platforms; the experience is coherent, but identity, evidence and exit remain captive to the runtime.

Range sensitivity0% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationfederated demand-and-supply network

A different purchasing world · illustrative protocol-enabled purchasing experience

Commerce becomes a routing network

YouPublish bounded intentSend a time-limited care coordination request without selecting a marketplace first.

NetRoute across networksGateways find compatible seller, logistics, financing and assurance participants.

BizReturn interoperable offersParticipants answer the same schema with inventory, terms, evidence and service capacity.

AICompose the outcomeRank whole outcomes rather than isolated products and preserve independent alternatives.

NetCoordinate fulfillmentCommitments and status updates travel across the selected participant graph.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:The selected bundle can combine inventory, service and physical execution from participants that never shared one marketplace owner.

Mechanism dossier · conditional scenario

Demand is broadcast into interoperable buyer, seller, logistics and assurance networks instead of sent to one seller agent.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

BecknONDCUCP discovery

Activation signal19%

2040 direct value pool$7.3B

Dominant in range draws0%

Human surfacea composed outcome from many network participants

Control remains hereChoose the routing and ranking policy, inspect participant provenance and exclude any network or seller class.

What could break it: Fragmented governance can make responsibility ambiguous when several participants jointly create a failed outcome.

All six mechanisms across 512 draws0% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2036
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Agents buy small pieces of digital work

Machine payments make data, search, verification and generation callable without a subscription or procurement cycle.

What purchasing feels like hereThe person approves a capped task; the assistant buys a few services and shows cost beside the resulting recommendation.

Conditional direct annual value pool$271M$137M–$493M sensitivity · based on metered machine services. Non-additive.

What must become true nextCapabilities can quote, authenticate, settle and return useful evidence cheaply.

How this path can stallMicro-payments add friction without proving that the work improved the outcome.

walled institutionThis path scales fastest inside integrated platforms; the experience is coherent, but identity, evidence and exit remain captive to the runtime.

Range sensitivity0% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationmachine-payable service graph

A different purchasing world · illustrative protocol-enabled purchasing experience

Everything becomes a paid capability

YouFund an outcome reservePlace a capped reserve behind the care coordination policy rather than prepaying one bundle.

AIDiscover paid capabilitiesFind machine-readable services for evidence, planning, capacity and execution.

NetOpen metered sessionsAuthorize capped sessions and stream settlement as useful work is verified.

AIRecompose continuouslyReplace underperforming services without discarding the completed work or policy.

YouAudit spend and resultSee cost by capability, final evidence and every automatically stopped session.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:The reserve is charged as verified work arrives, allowing the service graph to change without rebuilding a cart.

Mechanism dossier · conditional scenario

The commercial unit becomes a session, result or stream of work rather than a cart containing a finished product.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

MPPx402MCP

Activation signal13%

2040 direct value pool$3.1B

Dominant in range draws0%

Human surfacea reserve, live spend stream and result ledger

Control remains hereSet the reserve, per-service ceilings and stop conditions; review the live spend stream or terminate every session.

What could break it: Many individually cheap calls can create runaway spend, dependency chains or incentives to prolong the task.

All six mechanisms across 512 draws0% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2036
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Agents inherit platform reputation

Assistants can use marketplace ratings and escrow, but identity, ranking and dispute history remain trapped inside each platform.

What purchasing feels like hereThe person sees which platform guarantee backs the recommendation and must enter that platform to resolve failure.

Conditional direct annual value pool$1.2B$229M–$3.6B sensitivity · based on agent-executed GMV at a 1.2% scenario take. Non-additive.

What must become true nextAgent actions can be attributed to verified accounts and bounded payment instruments.

How this path can stallSynthetic reviews and account farming make inherited reputation less useful to machines.

walled institutionThis path scales fastest inside integrated platforms; the experience is coherent, but identity, evidence and exit remain captive to the runtime.

Range sensitivity0% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationregistry-and-escrow work market

A different purchasing world · illustrative protocol-enabled purchasing experience

Trust becomes the marketplace

AIPost a verifiable jobPublish the care coordination result, acceptance tests, deadline and escrow terms.

NetFilter identity and reputationResolve registered providers, attestations, prior outcomes and required bonds.

BizBid with evidenceProviders commit price, method, service terms and a bond against non-performance.

NetEvaluate the outcomeIndependent evaluators check delivery evidence and open a dispute window.

YouAccept or appealEscrow releases automatically inside policy; contested judgments return with the full record.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:Unknown providers can compete without requiring one platform to own identity, ranking, payment and dispute resolution.

Mechanism dossier · conditional scenario

Identity, reputation, escrow and independent evaluation let unknown agents form temporary markets around a job.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

ERC-8004ERC-8183A2A

Activation signal20%

2040 direct value pool$7.7B

Dominant in range draws0%

Human surfaceproof thresholds, bonds, evaluators and recourse

Control remains hereChoose reputation sources, minimum bonds, evaluators and appeal rules before the job is posted.

What could break it: Reputation can be gamed, evaluators can collude and on-chain evidence can preserve a wrong judgment permanently.

All six mechanisms across 512 draws0% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From the 2029 handoff to 2040

How this future becomes possible.

Follow the mechanism through bridge, formation, scale and infrastructure. Activation is model-derived support; milestones and purchasing moments are conditional narration.

largest modeled acceleration · 2036
conditional protocol-future path, not a forecast

bridge · 2029 · conditional path

Products expose a persistent identity

Digital links and event histories let a person or assistant retrieve authoritative product, ownership and service information after purchase.

What purchasing feels like hereThe person scans or asks the assistant about the asset; service history and compatible options replace a fresh web search.

Conditional direct annual value pool$113M$28.1M–$282M sensitivity · based on physically orchestrated GMV at a 3.0% scenario take. Non-additive.

What must become true nextProduct identifiers resolve to portable, owner-accessible lifecycle data.

How this path can stallManufacturers control the history, withhold repair data or abandon the resolver.

walled institutionThis path scales fastest inside integrated platforms; the experience is coherent, but identity, evidence and exit remain captive to the runtime.

Range sensitivity0% of 512 draws select this as the strongest mechanism at 2029; not a probability.

2040 destination · complete illustrative transaction

2040 · Care coordinationasset-to-service coordination loop

A different purchasing world · illustrative protocol-enabled purchasing experience

Physical objects become the buyers

ObjDetect a state changeThe asset records the care coordination condition with time, location, provenance and sensor evidence.

ObjCheck its service policyDetermine whether to observe, reserve capacity, request approval or stop operation.

NetFind compatible physical serviceQuery parts, technicians, facilities and logistics against the asset identity and history.

AIAuthorize the responseApply the owner’s mandate, preserve alternatives and bind the selected physical plan.

ObjVerify the new stateRecord completion evidence, update the twin and release payment only for the verified result.

Physical resultExisting care continues, provisional capacity is held and no new provider receives authority until a person decides.

What the person gets:The asset’s condition, service history and verified result form one continuing commercial record instead of separate shopping and maintenance journeys.

Mechanism dossier · conditional scenario

Assets publish their own condition and service history, allowing a vehicle, appliance, building or inventory item to initiate commerce.

These named protocols are architectural inspirations, not modeled winners or adoption forecasts.

GS1 Digital LinkEPCISdigital twins

Activation signal22%

2040 direct value pool$21.0B

Dominant in range draws0%

Human surfacea living asset timeline with revocable operating policy

Control remains hereControl which sensors and service classes may initiate action, require approval for irreversible work and revoke the asset mandate.

What could break it: Bad sensor data, compromised twins or manufacturer-controlled histories can trigger unnecessary work or lock owners into captive service.

All six mechanisms across 512 draws0% selectedSelection share is sensitivity coverage from unanchored ranges, not a probability that this protocol future occurs.

From one assistant to an agent ecology

Eight forms can coexist—and diverge.

strongest form outlined
physical implications authored

interaction form2029203220362040

Site referralPhysical fulfillment remains downstream of a conventional order.

84%

72%

59%

50%

Delegated checkoutThe seller fulfills an agent-authored order rather than a person-built cart.

13%

20%

25%

28%

Bilateral agent tradeInventory, service capacity and delivery can be reserved machine-to-machine.

13%

Terms marketThe physical plan changes before checkout because timing and capacity are part of the bargain.

Standing outcome policySupply, repair, transport or care coordination continue without a new shopping session.

Collective procurementNeighborhoods, fleets or firms reshape routes, inventory and capacity through coordinated demand.

Commissioned supplyProduction schedules, components and facilities respond to intent rather than forecasted shelf demand.

14%

Embodied orchestrationThe purchasing interface becomes a live map of actions, safety boundaries and revocable authority.

15%

23%

Delegation frontier · published path

Where autonomy stops depends on the need.

engine-selected interface + full range coverage · unanchored priors · illustrative UI

commerce arena2029203220362040

Household replenishmentexplore this needapproval100% selectedapproval30% selectedpolicy61% selectedpolicy60% selected

Everyday mobilityexplore this needapproval100% selectedapproval88% selectednegotiate62% selectednegotiate60% selected

Complex travelexplore this needapproval100% selectedapproval100% selectedapproval96% selectedapproval69% selected

Care coordinationselected lensapproval100% selectedapproval100% selectedapproval96% selectedapproval69% selected

approvalnegotiatepolicyambient

Change the assumptions behind this story

Server simulationEvery submitted counterfactual is recomputed by the deterministic horizon engine; the URL preserves the complete run state.

Long-horizon counterfactual · structural assumptions

Change what becomes possible.

These inputs rerun the selected path from its 2029 scenario state. They do not alter the near-term engine or any dated observation.

published preset

Commerce arena

High-consequence services where continuity, eligibility and personal fit stay salient. The arena changes interface thresholds and human attention, not the aggregate market path.

repeatability55%reversibility25%consequence90%preference load75%

unanchored scenario prior · modeled category lens

commercial capability learning / year17%8% → 24% · unanchored research range

persistent-agent cost decline / year26%15% → 45% · unanchored research range

seller-agent diffusion / year11%6% → 24% · unanchored research range

portable authority and receipts80%10% → 90% · unanchored research range

liability clarity82%10% → 90% · unanchored research range

fraud and manipulation pressure / year13%8% → 35% · unanchored research range

physical execution diffusion / year9%2% → 18% · unanchored research range

portable pooled mandates28%5% → 75% · unanchored research range

adaptive supply diffusion / year5%2% → 18% · unanchored research range

Only the selected structural path is replaced; the comparison lines remain the published presets.

Benchmark the result

Three regimes become reference points, not destiny.

The reader-built path can combine mechanisms that the old labels kept apart. Coordinated-open, fragmented-open and walled presets remain below as transparent comparison anchors for concentration and interoperability.

Coordinated open rails

agent markets

54% of commerce has a machine on both sides; 23% runs continuously against a policy. Conditional 2040 p10–p90: 18%–48% machine counterparties.

Fragmented open web

delegated commerce

22% of commerce has a machine on both sides; 6% runs continuously against a policy. Conditional 2040 p10–p90: 12%–34% machine counterparties.

Walled gardens

delegated commerce

16% of commerce has a machine on both sides; 4% runs continuously against a policy. Conditional 2040 p10–p90: 8%–24% machine counterparties.

Machine counterparties · published presets with p10–p90 structural sensitivity

CoordinatedFragmentedWalled

Structural sensitivity ribbons. Each p10–p90 band covers 512 paired, seeded combinations of the nine disclosed unanchored structural ranges, conditioned on that branch’s published 2029 handoff. The line is the published preset; the ribbon is not a confidence interval or probability. A preset can sit outside its ribbon because the coverage design spans every range stratum rather than centering on that preset.19

Beyond the site. Embodied agents, autonomous mandates, agent trust registries and transaction-lifecycle protocols demonstrate partial building blocks—not adoption rates or a complete physical economy. Pooled demand, commissioned supply and embodied orchestration remain disclosed structural possibilities with unanchored priors.20 21 22 23

Research appendix: new actors and assumptions

Actors the 2029 model does not contain

The cast changes with the market.

These actors enter only when the structural layer reaches the relevant epoch.

Merchant capabilityenters 2032exposes discovery, checkout, fulfillment and support as callable operations without needing a human-facing storefrontnew actor

Seller agententers 2032sets price, inventory exposure, service terms and which buyers to trustnew actor

Authority networkenters 2032proves what an agent may spend, disclose, sign and delegate onwardnew actor

Open commerce routerenters 2032routes portable intent across catalogs, logistics and payment providers without owning the buyer relationshipnew actor

Agent market operatorenters 2036matches machine demand and supply while setting access and ranking rulesnew actor

Assurance networkenters 2036underwrites disputes, provenance, delivery and adversarial behaviornew actor

Physical operator agententers 2036commits robots, vehicles, workers and facilities to machine-negotiated physical outcomesnew actor

Metered capabilityenters 2036quotes and settles small units of digital or physical work as they are consumednew actor

Trust and escrow registryenters 2036makes identity, reputation, job state and conditional release portable between unfamiliar counterpartiesnew actor

Demand coalitionenters 2036pools bounded mandates so households or organizations bargain without surrendering individual constraintsnew actor

Resource agententers 2040continuously allocates money, energy, compute, inventory or capacity against a policynew actor

Asset twinenters 2040turns a machine, package, building or product lifecycle into an authenticated source of commercial intentnew actor

Production agententers 2040configures or commissions supply after machine demand forms, before a finished product necessarily existsnew actor

Research hooks

What the next evidence pass must update.

capabilityLearningannual improvement in agents completing multi-step commercial workprior range 0.08–0.24unanchored

inferenceCostDeflationannual decline in the cost of persistent commercial agencyprior range 0.15–0.45unanchored

sellerAgentDiffusionannual diffusion of seller-side pricing, inventory and negotiation agentsprior range 0.06–0.24unanchored

identityPortabilityhow portable identity, authority, reputation and receipts become across runtimesprior range 0.1–0.9unanchored

liabilityClarityclarity about who bears loss when an autonomous transaction goes wrongprior range 0.1–0.9unanchored

adversarialPressureannual pressure from fraud, collusion, manipulation and synthetic supplyprior range 0.08–0.35unanchored

physicalExecutionDiffusionannual diffusion of agent-addressable robots, vehicles, workers and facilitiesprior range 0.02–0.18unanchored

collectiveMandatePortabilityability to pool bounded purchasing mandates across people or organizationsprior range 0.05–0.75unanchored

adaptiveSupplyDiffusionannual diffusion of supply agents that can configure or commission capacity from demandprior range 0.02–0.18unanchored

Enter one illustrative agent-to-agent transaction →

Commerce becomes an operating layer.

Buying stops being one kind of event.

Choose what becomes true—then live through the result.

The future is a recipe, not a regime.

The mandate utility network

Coordinated open rails

One need moves from request to operating policy.

Checkout compatibility exposes the seller gap.

One intent, one approval

The agent prepares the visit; the person chooses the provider.

The structural handoff begins as an approval bundle.

Mandates become portable credentials.

One intent, one approval

Administrative terms negotiate; care decisions do not.

approval bundle remains visible while its mandate deepens.

Agent auctions begin clearing commercial terms.

Terms become negotiable

Care logistics become continuous while clinical authority stays human.

The interface moves from approval bundle to negotiation room.

Standing policies replace more checkout sessions.

The purchase becomes a policy

The most capable agent is visible in the moments it refuses to act.

The interface moves from negotiation room to policy console.

A dollar can be influenced, executed, retained—or captured as infrastructure.

Six futures can grow from the same need.

How this future becomes possible.

Assistants learn to carry merchant state

Checkout becomes an embeddable capability

Merchant capabilities compose into journeys

The assistant becomes a universal storefront layer

The assistant is the storefront

How this future becomes possible.

One-use authority becomes legible

Authority travels across agents and merchants

Standing mandates replace repeated approval

Permission becomes the commercial primitive

Permission replaces checkout

How this future becomes possible.

Adapters translate fragmented commerce

Portable intent reaches federated discovery

Networks compose supply, logistics and assurance

Commerce behaves like a routing network

Commerce becomes a routing network

How this future becomes possible.

Agents buy small pieces of digital work

Capped sessions replace API subscriptions

Service graphs recompose while the outcome runs

The economy exposes paid capabilities, not finished carts

Everything becomes a paid capability

How this future becomes possible.

Agents inherit platform reputation

Identity and outcome history become portable

Temporary markets form around verifiable jobs

Trust infrastructure becomes the marketplace

Trust becomes the marketplace

How this future becomes possible.

Products expose a persistent identity

Assets diagnose and prepare a response

Assets reserve physical capacity inside policy

Physical objects initiate and verify commerce

Physical objects become the buyers

More agents create a new coordination problem.

Eight forms can coexist—and diverge.

Where autonomy stops depends on the need.

Fragmented open web

One need moves from request to operating policy.

Open protocols produce an adapter economy.

One intent, one approval

The agent prepares the visit; the person chooses the provider.

The structural handoff begins as an approval bundle.

Credential translation becomes its own market.

One intent, one approval

Administrative terms negotiate; care decisions do not.

approval bundle remains visible while its mandate deepens.

Seller agents become negotiating counterparties.

One intent, one approval

Care logistics become continuous while clinical authority stays human.

approval bundle remains visible while its mandate deepens.

Persistent agents remain bounded by reconciliation debt.

Terms become negotiable