Obenan Research · Flagship briefing · May 2026
Prepaid local services: a practical wedge from agentic discovery to agentic payment
Networks, model providers, and platform companies are building credible authorization, tokenization, intent, checkout, and protocol layers for agentic commerce. That work governs who is allowed to pay. It does not, by itself, govern whether the merchant on the other side of a specific local commitment is real, present, available, and able to honor it in the moment. Bounded local commitments are where that question becomes answerable before money moves.
Published May 1, 2026
Operator takeaway
Treat bounded commitment objects, not generic local SMBs, as the agent-payable unit. Run two pilot lanes in parallel: a payment-readiness lane (parking, ticketed experiences, classes) to prove the rails, and a merchant-truth lane (beauty, personal services, appointment-led wellness) to prove the upstream readiness layer earns its place.
- Published
- May 2026
- Format
- 27-page PDF
- Sources
- 15 primary citations
- Pilot lanes
- 2 in parallel
Open research briefing · 27-page PDF
Executive scan
Three questions this briefing answers, in order
Read these three answers in order. They frame the rest of the document and isolate the part of the stack Obenan is responsible for.
What local-service commitments can agents safely pay for?
Bounded commitment objects: discrete, named, time-stamped, fulfillable promises with a clear capacity, price, cancellation rule, and merchant confirmation step. Parking bays, timed-entry tickets, class seats, spa appointments, restaurant deposits, pickup orders, and rental windows fit. Generic local SMBs do not.
Where does payment infrastructure stop and merchant truth begin?
Networks, tokenization, mandates, agent wallets, and machine-payment primitives govern who is allowed to pay and what they are allowed to pay for. They do not establish that the slot, ticket, price, or table was actually available at the moment of commitment. That is merchant-side truth, upstream of payment execution.
Which pilot lanes prove payment readiness versus merchant readiness?
Lane A (parking, ticketed experiences, classes) shows the rails work end to end where the bounded object is already digital. Lane B (beauty, personal services, appointment-led wellness) shows why an upstream merchant-truth layer earns its place where availability is volatile and the unit is fragmented.
Five takeaways
Five takeaways
For the reader with two minutes.
- 01
Public agentic infrastructure is solving authorization, not commitment
Registered agents, network tokens, signed user mandates, verified intent, agent wallets, and machine-payment primitives govern who is allowed to pay. They do not, by themselves, govern whether the merchant on the other side can honor the specific commitment being paid for.
- 02
A local service is not a SKU
It is a time-bound, capacity-bound, location-bound, operationally volatile, often policy-heavy promise made by a small operator, rarely represented by one structured catalog.
- 03
The wedge is bounded, not broad
The category that is ready is narrower and well defined: prepayable, reservable, ticketable, or deposit-backed services with a single, named, time-stamped commitment object that an agent can verify before money moves.
- 04
Two pilot lanes, not one
A payment-readiness lane (parking, ticketed experiences, classes) shows the rails work end to end. A merchant-truth lane (beauty, personal services, appointment-led wellness) shows why an upstream readiness layer earns its place in the stack.
- 05
Obenan sits upstream of payment execution
No money movement, no tokenization, no authorization, no checkout, no acquiring, no settlement. The work is merchant-side: identity, eligibility, freshness, validation, evidence, and committability before the payment layer takes over.
What changed
What changed in the past forty days
The execution layer of agentic commerce hardened materially in April and early May 2026. The pattern across releases is the same: the public infrastructure is moving from launch rhetoric into operating contracts.
Visa public rulebook adds Agentic Platform Requirements
Visa published rulebook v2026-04-18 with a new section 4.1.24 Agentic Platform Requirements, formalizing enrollment, consent, merchant-policy acknowledgement, post-transaction evidence, repeated-transaction consent, and 120-day order confirmation requirements that include cancellation and guaranteed-reservation disclosure.
Universal Commerce Protocol cuts a versioned release
The Universal Commerce Protocol cut release v2026-04-08 on April 9, 2026, adding cart capability, catalog search and lookup, signed request and response envelopes, eligibility claims, authorization and abuse signals, and explicit totals and disclosure contracts. The release explicitly distinguishes provisional catalog truth from authoritative checkout truth.
Agentic Commerce Protocol publishes a new revision
The Agentic Commerce Protocol published version 2026-04-17, adding a Feed API, a Cart capability, delegated consumer-authentication for 3DS2, a public well-known discovery document with version and capability signaling, and explicit transport signaling for both REST and MCP.
Stripe expands its agentic commerce position
Stripe at Sessions 2026 broadened merchant access through the dashboard, added Google as a downstream discovery and checkout route via AI Mode and the Gemini app, and introduced Link agent wallets and Machine Payments Protocol support for programmatic agent payments.
Experian announces an agent identity and trust layer
Experian introduced Agent Trust, with Human-to-Agent Binding, an Agent Trust Token, and a dynamic Agent Registry, positioned as complementary to network and edge layers.
Adyen plans to acquire Talon.One for pre-payment decisioning
Adyen announced its plan to acquire Talon.One for EUR 750 million, explicitly tying the deal to agentic-commerce readiness and to real-time pre-payment decisioning across cart and offer state.
The center of gravity in this stack is consolidating around authorization, tokenization, identity, intent, agent wallets, and pre-payment decisioning. The merchant-side question of whether a specific local commitment can actually be honored remains under-specified.
Why prepaid local services are different
A local service is a promise, not a SKU
For agentic commerce, the most important property of a purchasable object is whether its truth can be checked before money moves. A digital product carries that truth in a SKU, a price, a stock level, and a return policy. A local service carries it in something messier.
A local service is time-bound, capacity-bound, location-bound, operationally volatile, often policy-heavy, and frequently distributed across booking platforms, scheduling tools, and informal channels. Treat a salon appointment, a yoga class, or a fixed-window pickup like a product on a feed and the agent will, on average, get the result that pattern earns: a payment that cleared against a commitment the merchant could not honor.
Time-bound
The same service at six in the evening is not the same service at noon.
Capacity-bound
Most categories operate against a small, hard limit (seats, bays, chairs, tables).
Location-bound
Fulfillment happens in one place, often by one person.
Operationally volatile
Staff call out, equipment breaks, weather shifts, cancellations create gaps that may not refill.
Policy-heavy
Cancellation, deposit, late-arrival, and modification rules vary by category, region, and operator.
Rarely a single catalog
A merchant's truth lives across multiple booking systems, scheduling tools, and human routines.
Bounded commitment objects
The bounded commitment object
A prepaid local service becomes agent-ready when the thing being purchased is bounded enough for an agent to reason about and the merchant operationally able to confirm. The shape of the object matters more than the vertical it sits in. A bounded commitment object is a discrete, named, time-stamped, fulfillable promise. It has a clear start and end, a clear capacity, a clear price, a clear set of conditions under which it can be honored or canceled, and a clear merchant confirmation step. It is, in effect, a small contract that an agent can verify before money moves.
A parking bay: a specific space, in a specific garage, for a specific window of time, at a clearly published rate.
A timed-entry ticket: a defined entry window at a museum, attraction, or experience venue, with fixed price and fixed quantity.
A class seat: one position in a yoga, cooking, or fitness session, with known capacity and a known cancellation rule.
A spa or wellness appointment: a defined treatment, a defined duration, a defined practitioner, with deposit or prepayment behavior already common in the category.
A restaurant deposit: a reservation backed by a card hold or partial prepayment, used by major restaurant platforms as a no-show mitigation tool.
A pickup order: a short list of items, a confirmed total, a defined pickup window, against a merchant that already handles the same flow for human-driven mobile orders.
A rental window: a specific piece of equipment, a specific time block, a defined deposit, a defined damage policy.
The same merchant may produce some commitments that are agent-ready and others that are not. The line is drawn around the object, not around the operator.
Where payment infrastructure stops
Where payment infrastructure stops, and why merchant-side truth still has to land
Public agentic payment infrastructure can verify the user, the agent, and the payment mandate. It can govern who is allowed to pay and what they are allowed to pay for. Live, controlled implementations have already been demonstrated. The trajectory is toward a payment layer that knows who is allowed to pay, what they are allowed to pay for, and how much.
That work answers the credential question. It does not, by itself, answer the merchant-readiness question. Tokenization protects credentials. Mandates bind user instructions to agent actions. Neither establishes that the slot, the ticket, the price, or the table was actually available at the time the commitment was made.
That gap is upstream of payment execution and downstream of discovery. It is also distinct from the AI Visibility and Discoverability question of whether a merchant is findable on a given AI surface. Discoverability asks whether the agent can see the merchant. Committability asks whether the merchant can honor a specific named commitment right now. Both matter. They are not the same layer.
Discovery
An AI agent finds the merchant on a given surface. This question is whether the agent can see the merchant at all.
AI surfaces, agents, search
Merchant truth
Identity, eligibility, freshness, validation, evidence, and committability of a specific commitment. Whether the slot, ticket, price, or table is actually honorable right now.
Obenan
Payment execution
Authorization, tokenization, mandates, agent wallets, machine-payment primitives, checkout, and settlement. Who is allowed to pay and how money moves.
Networks, PSPs, acquirers, wallets
Three layers, one boundary. Obenan operates upstream of payment execution and downstream of discovery.
Two pilot lanes
Two pilot lanes, not one
The temptation in early infrastructure is to pick one vertical and force the entire argument through it. That is a mistake here. The two questions a pilot must answer are different enough that no single category can answer both without one of the answers becoming theoretical.
Lane A · Payment readiness
Categories where the rails can already prove themselves
Categories where the bounded commitment object is already digital, the inventory is already managed by mature systems, and the unit economics of a single transaction are already well understood. Candidates: ticketed experiences, classes, parking. Each produces a clean object an agent can purchase. Each interacts with infrastructure that already supports digital payment, deposit handling, and cancellation logic. Lane A answers a focused question: when an agent reaches the point of commitment in a category that is already mostly digital, does the end-to-end flow work as designed.
Lane B · Merchant truth
Categories where the merchant side is the hardest part
Beauty and personal services lead the list. Spa, wellness, and other appointment-led categories follow closely. Availability is genuinely volatile, practitioner skills are genuinely differentiated, cancellation behavior is genuinely category-specific. Lane B answers a different question: in a category where the merchant side is fragmented, can a dedicated readiness layer produce evidence strong enough to make the commitment safe to proceed.
The two lanes are complementary. Lane A shows the rails work. Lane B shows why upstream merchant infrastructure is needed before agents can commit on the user's behalf at scale.
Evidence discipline
Evidence discipline
This briefing is a synthesis of public sources. Each claim is traceable. Three layers of evidence are used and labeled.
Official and primary
Company product pages, public specifications, public press releases, government statistical sources, and standards-body release tags. This tier carries the body claims.
High-quality industry reports
Commercial market reports and consultancy publications. Used as category-scale anchors only, not as agentic-commerce market sizing.
Vendor primary, self-interested
Company blog posts and product pages that describe their own outcomes. Labeled when used. Treated as direction, not as comparative proof.
Observed: every public release, specification, and rulebook cited above. Inferred: that the merchant-side question of whether a specific local commitment can be honored remains under-specified across the current public stack.
Obenan point of view
Obenan's position
Obenan helps physical merchants become safely discoverable and committable by AI agents before payment execution begins. The work is upstream: identity, eligibility, freshness, validation, evidence, and committability. Obenan does not move money, tokenize cards, authorize payments, act as a payment service provider, act as an acquirer, build checkout, or settle transactions.
A clean agentic ecosystem will need both layers. Networks, payment service providers, acquirers, agent wallets, and machine-payment primitives govern authorization. The merchant layer must govern truth at readiness. Prepaid local services are the category where the distinction stops being theoretical.
Non-claims and boundaries
What this briefing does not claim
The argument above is bounded. The list below is the explicit set of claims this briefing does not make.
- 01
This is not a claim that all local commerce is ready for agentic payment.
- 02
This is not a claim that prepaid local services are a proven broad commercial rollout.
- 03
This is not a Mastercard approval, integration, partnership, Start Path, pilot, or developer-program claim.
- 04
This is not a Visa claim or any private-network claim. References to public Visa rules are public rulebook references only.
- 05
This is not a claim of integration with Mastercard Agent Pay, Mastercard Verifiable Intent, Visa Intelligent Commerce Connect, American Express ACE, the Universal Commerce Protocol, the Agentic Commerce Protocol, AP2, Experian Agent Trust, Stripe Agentic Commerce Suite, or Adyen real-time decisioning.
- 06
This is not a payment-service-provider, acquirer, checkout, tokenization, or payment-execution claim. Obenan is not in that lane.
- 07
This is not legal, regulatory, or payment-network rules advice. Operators should consult their own counsel and their network or processor of record.
- 08
This briefing does not publish private partner names, private meeting state, or NDA material. The argument is built from public sources only.
Related Obenan Research
Continue the thread
Discuss the research
This briefing exists because the merchant side of agentic commerce still has to be built. Conversations with operators, networks, payment service providers, acquirers, vertical platforms, and merchant-system integrators are the most useful next step.
Research briefing · 27-page PDF
Sources
Every claim in this briefing is anchored to a public source. The list below preserves the same chronology and source labels used in Obenan's underlying research pack.
Primary public sources
- 1.Visa Core Rules and Visa Product and Service Rules (public rulebook)usa.visa.com · April 18, 2026 · Checked April 24, 2026
Public rulebook PDF including new section 4.1.24 Agentic Platform Requirements; the public-rules anchor for the network side of the boundary.
- 2.Universal Commerce Protocol release v2026-04-08github.com · April 9, 2026 · Checked April 21, 2026
Open-protocol release tag with cart capability, catalog search and lookup, signed envelopes, eligibility claims, authorization and abuse signals, and totals and disclosure contracts.
- 3.UCP catalog specification (live)ucp.dev · April 9, 2026 · Checked April 21, 2026
Live UCP catalog specification distinguishing provisional catalog truth from authoritative checkout truth.
- 4.UCP checkout REST specification (live)ucp.dev · April 9, 2026 · Checked April 21, 2026
Live UCP checkout REST specification carrying the authoritative checkout contract.
- 5.Agentic Commerce Protocol changelog 2026-04-17github.com · April 17, 2026 · Checked April 24, 2026
ACP changelog with Feed API, Cart capability, delegated 3DS2 authentication, well-known discovery document, and REST plus MCP transport signaling.
- 6.Stripe Sessions 2026 newsroomstripe.com · April 29, 2026 · Checked May 1, 2026
Vendor newsroom summarizing Sessions 2026 product expansion across agentic commerce, dashboard merchant access, and Link agent wallets.
- 7.Stripe agentic commerce documentationdocs.stripe.com · April 29, 2026 · Checked May 1, 2026
Vendor product documentation for Stripe's agentic commerce surface and Machine Payments Protocol support.
- 8.Experian Agent Trust press releaseexperianplc.com · April 30, 2026 · Checked May 1, 2026
Press release introducing Human-to-Agent Binding, an Agent Trust Token, and an Agent Registry.
- 9.Adyen acquires Talon.One press releaseadyen.com · April 23, 2026 · Checked May 1, 2026
Press release for the planned EUR 750 million acquisition framed around real-time pre-payment decisioning and agentic-commerce readiness.
- 10.Mastercard Agent Pay public product pagemastercard.com · Public evergreen · Checked May 1, 2026
Public product page describing Mastercard Agent Pay and surrounding agentic commerce capabilities.
- 11.Verifiable Intent specification (public draft v0.1)verifiableintent.dev · Public draft · Checked May 1, 2026
Public draft specification for Verifiable Intent.
- 12.Mastercard and Santander first live end-to-end AI agent payment in Europe (controlled)mastercard.com · March 2, 2026 · Checked May 1, 2026
Press release announcing a controlled live end-to-end agent-executed payment in Europe; cited as a context point, not a broad commercial rollout.
- 13.OpenAI Buy It in ChatGPT (Instant Checkout via Agentic Commerce Protocol)openai.com · Public evergreen · Checked May 1, 2026
Public product page for the OpenAI Instant Checkout surface implemented over the Agentic Commerce Protocol.
- 14.OpenAI product feed specificationdevelopers.openai.com · Public evergreen · Checked May 1, 2026
Public OpenAI product feed specification supporting the Buy It in ChatGPT surface.
- 15.Eurostat e-commerce statisticsec.europa.eu · Public evergreen · Checked May 1, 2026
Government statistical source for European e-commerce participation; used to anchor merchant digitization context.
High-quality public industry sources (used as category-scale anchors only)
- 16.Fortune Business Insights, Health and Fitness Club Marketfortunebusinessinsights.com · 2024 estimate · Checked May 1, 2026
Industry market report used as a category-scale anchor only, not as agentic-commerce market sizing.
- 17.Grand View Research, Professional Beauty Services Market Reportgrandviewresearch.com · 2023 estimate · Checked May 1, 2026
Industry market report used as a category-scale anchor only, not as agentic-commerce market sizing.
- 18.Grand View Research, Medical Spa Marketgrandviewresearch.com · 2024 estimate · Checked May 1, 2026
Industry market report used as a category-scale anchor only, not as agentic-commerce market sizing.
- 19.Research and Markets, Tours and Activities Reservations Globalresearchandmarkets.com · 2024 estimate · Checked May 1, 2026
Industry market report used as a category-scale anchor only, not as agentic-commerce market sizing.
- 20.GM Insights, Parking Reservation System Marketgminsights.com · 2024 estimate · Checked May 1, 2026
Industry market report used as a category-scale anchor only, not as agentic-commerce market sizing.
- 21.SpotHero, one billion dollars cumulative parking reservations milestonespothero.com · Vendor self-reported milestone · Checked May 1, 2026
Vendor self-reported milestone; treated as direction, not as comparative proof.
- 22.OpenTable, three proven payment strategies to reduce no-showsopentable.com · Vendor evergreen guidance · Checked May 1, 2026
Vendor evergreen guidance on prepayment and deposit strategies for restaurant reservations.
- 23.Tock, eliminate no-shows: three Tock toolsexploretock.com · Vendor self-reported · Checked May 1, 2026
Vendor self-reported tooling and outcomes for restaurant deposit and prepayment patterns.
- 24.McKinsey, the agentic commerce opportunitymckinsey.com · 2026 publication · Checked May 1, 2026
Consultancy report on agentic commerce trajectory; used as a directional anchor, not as commercial proof for prepaid local services.