top of page
Website Quantum Payments (1).png

The Agentic Shortlist Closes in October: Payment Readiness Is the Real Bottleneck

Writer: Quantum Payments
Quantum Payments
4 days ago
10 min read

Executive summary

Agentic commerce has moved from sandbox demonstrations into production payment flows. The industry is now debating which agent payments protocol will prevail: OpenAI and Stripe’s ACP, Google’s AP2, Visa’s Trusted Agent Protocol, Mastercard Agent Pay, EMVCo’s emerging framework or the Agent Payments Group’s MPP.

That debate matters to infrastructure providers. It is not the most urgent question for merchants.

The immediate issue is that AI agents are already discovering products and narrowing holiday shortlists in October, while many merchants have not made their payment stack ready to be bought from by a machine. Discoverability is advancing faster than transaction readiness. The discovery-to-authorisation gap is becoming the real bottleneck.

The merchants that close this gap now can participate in the first agentic holiday demand cycle. Those that wait until January will be measuring the demand they failed to convert.

The numbers show a discovery-to-authorisation gap

The latest research makes the timing difficult to ignore.

Shopify’s first agentic holiday season analysis says 65% of shoppers plan to use AI for at least one shopping task this season. It also warns that agents can finalise purchase shortlists as early as October. These agents do not assess a merchant through banner advertising or brand familiarity alone. They use structured information such as product titles, dimensions, materials, pricing, availability and policies.

American Express reaches a similar conclusion from the merchant side. Its Business Playbook for Agentic Commerce, released on 6 October 2026, found that:

  1. 62% of merchants expect AI agents to have a moderate or significant impact on their business within the next year.

  2. 83% see agentic commerce as an opportunity.

  3. Only 41% are experimenting with or implementing it.

  4. 66% of US consumers are likely to use AI agents for price research.

  5. 61% are likely to use agents for recommendations.

  6. Only 15% are comfortable allowing an agent to buy high-value items such as electronics.

This is not simply a technology adoption gap. It is a commercial hand-off problem. Agents can increasingly find the merchant and recommend the product, but the payment environment may still be unable to identify the agent, verify delegated intent, interpret machine-readable returns policies or assemble evidence if the transaction is disputed.

Agent-assisted and agent-delegated commerce are not the same

Federal Reserve Governor Christopher J. Waller’s Sibos speech, “Payments in the Age of AI Agents”, provides a useful commercial distinction.

In agent-assisted commerce, the agent researches, compares and recommends. The consumer remains in control and completes the payment. This resembles high-intent referral traffic. The merchant’s priority is to be legible, competitive and easy to convert.

In agent-delegated commerce, the consumer gives an agent authority to transact within defined constraints. Those constraints may include a budget, category, supplier, quantity, time window or recurring-payment rule. The agent may complete the purchase without the consumer reviewing the final checkout screen.

That changes what the payment request must carry. A delegated transaction needs more than cardholder authentication. It needs evidence that:

  • the agent is identifiable;

  • the agent was authorised by the customer;

  • the purchase sits within the delegated scope;

  • the merchant’s product, price and policy matched the instruction; and

  • the transaction can be reconstructed after the event.

The difference is commercially important. Assisted traffic should be measured like a new acquisition and conversion channel. Delegated traffic must also be managed like a policy-controlled payment stream.

The rails are moving faster than merchant operations

This is no longer a purely theoretical problem.

In June 2026, Worldline, ING and Mastercard completed a live end-to-end agentic payment in production. Worldline has also documented live agent-driven payments across Visa and Mastercard environments.

On 30 September, Mastercard announced that its Agent Pay probability scoring was live for testing in the United States. The score estimates the likelihood that a transaction was initiated by an AI agent. It sits alongside identity verification from Skyfire and web-signal analysis from Cloudflare.

That score is an intelligence layer, not a pass-or-fail gate. Its value depends on what issuers, acquirers and merchants do with the signal. If a merchant’s fraud model treats machine-speed behaviour as automatically suspicious, the result may be a lower payment success rate even when the agent is legitimate.

Mastercard and Google’s Verifiable Intent framework addresses a different part of the problem. Its open specification uses cryptographic delegation chains to bind a human-approved scope to an agent’s actions. It is protocol-agnostic and designed to work with ecosystems such as AP2 and UCP.

EMVCo’s agentic payments work is also focused on consumer intent, agent identification and transaction-level signals.

The direction is clear: the industry is building trust signals. The merchant question is whether those signals can be absorbed without creating a separate integration for every AI platform.

The standards war is a distraction for merchants

The emerging landscape is becoming what PYMNTS described on 1 October as “the alphabet soup of agentic standards”. ACP, AP2, TAP, UCP, Agent Pay, EMVCo and MPP may all have important roles. But merchants should not place their operational roadmap on a single protocol bet.

The practical response is protocol abstraction.

A merchant should be able to maintain one structured catalogue, one policy layer, one risk framework and one reconciliation process while a payment service provider or orchestration layer manages the changing interfaces above it.

Merchant requirement

Human checkout assumption

Agentic commerce requirement

Product information

Designed for visual browsing

Structured for machine comparison

Consent

Captured at checkout

Delegated before or during the transaction

Authentication

Proves the payer’s identity

Proves identity and agent authority

Fraud controls

Calibrated to human behaviour

Calibrated to agent patterns and velocity

Returns and refunds

Written for customers to read

Machine-readable and enforceable

Disputes

Order and payment records

Intent,scope,agent identity and execution evidence

Reporting

Channel and payment method

Agent traffic,protocol,approval and outcome

This is where payment orchestration becomes more than a routing conversation. It becomes an abstraction layer between fast-changing agent protocols and stable merchant operations.

Why low-risk baskets will arrive first

Consumer hesitation is highest where value-at-risk is highest. Only 15% of surveyed US consumers are comfortable letting an agent buy high-value electronics. That suggests the first scaled use cases will be low-risk, high-frequency and rules-based:

  • replenishment orders;

  • household essentials;

  • approved business procurement;

  • subscriptions;

  • repeat purchases; and

  • purchases within tight price and category limits.

These baskets are commercially attractive because they are predictable. They are also ideal for testing delegated intent, agent-aware fraud rules and automated reconciliation.

The risk is that a merchant treats agentic checkout as a front-end feature while leaving the back office unchanged. If the agent buys the wrong variant, uses an expired policy or exceeds a budget, the dispute will not be resolved by a smoother button. It will be resolved by evidence.

That evidence should connect the customer’s instruction, the agent identity, the applicable constraints, the catalogue version, the authorisation result, the fulfilment record and the refund or dispute decision.

Quantum Payments’ previous analysis on AI fraud detection and payment success rate and the frictionless authentication paradox is relevant here. Agentic payments add another dimension to the approval decision. The objective is not maximum friction or maximum automation. It is accurate, explainable decisioning.

The merchant readiness checklist

Merchants preparing for the October shortlist should prioritise the following:

  1. Identify the agent. Capture the agent platform, agent instance and relevant transaction signal where available.

  2. Record delegated intent limits. Store budgets, categories, suppliers, recurrence terms and time windows in an enforceable format.

  3. Make returns and refund policies machine-readable. Expose eligibility windows, conditions, fees and process steps in structured data.

  4. Create dispute evidence. Preserve the instruction, delegation chain, product data, price, authorisation result, fulfilment record and customer communications.

  5. Introduce agent-aware fraud rules. Distinguish authorised machine behaviour from automated attack patterns rather than blocking all high-velocity traffic.

  6. Report payment success rate by agent traffic. Separate assisted, delegated, human and unknown-origin transactions.

  7. Reconcile the complete transaction lifecycle. Connect order, authorisation, capture, refund, chargeback and settlement records.

  8. Map liability before launch. Define who is responsible when an agent acts outside scope, buys the wrong item or creates an unauthorised dispute.

A merchant does not need to predict which protocol wins. It needs an operating environment that can accommodate the winners.

What Quantum Payments’ position means

Quantum Payments is built around a unified commerce environment spanning online payments, in-person payments, embedded payments, automated workflows and payment orchestration. That matters because agentic commerce does not stop at checkout.

The merchant needs structured product and policy data at the front, agent-aware fraud and authorisation controls in the middle, and reconciliation, accounting and business intelligence at the back.

The Quantum Payments features platform brings these capabilities into one modular environment. It is designed to support rapid deployment, secure payment processing, omnichannel operations and automation from checkout through reconciliation.

The strategic principle is simple: do not make the merchant rebuild its commerce stack every time an AI platform introduces a new agent payments protocol.

The October shortlist is already forming. Discoverability is no longer the main question. The merchants that win the first agentic holiday cycle will be the ones that can turn a machine-readable recommendation into an authorised, fulfilable and defensible transaction.

Frequently asked questions

What is agentic commerce?

Agentic commerce is commerce in which an AI agent searches, compares, recommends or purchases products and services on behalf of a consumer or business.

What is the difference between agentic payments and ordinary online payments?

Ordinary online payments generally authenticate a payer and authorise a transaction. Agentic payments must also establish that an AI agent was authorised to act, understand the scope of that authority and preserve evidence of how the transaction was executed.

Do merchants need to choose one agent payments protocol?

No. Merchants should prioritise protocol abstraction through a PSP or payment orchestration layer. Their core catalogue, policy, risk and reconciliation systems should remain independent of any single protocol.

Why are agent-aware fraud controls important?

Fraud systems trained on human behaviour may misinterpret legitimate machine-speed activity as suspicious. Agent-aware controls help distinguish authorised agents from malicious automation while protecting payment success rate.

What should merchants do before the holiday season?

Start with structured product and policy data, agent identification, delegated intent limits, dispute evidence and reporting that separates agent-originated traffic from human traffic. These foundations are useful regardless of which protocol becomes dominant.

Sources

Publishing and social handover

Publishing instruction

  • Schedule in Wix: Thursday 8 October 2026 at 7:30am AEST.

  • Wix slug:agentic-commerce-payment-readiness-bottleneck

  • Blog URL:https://www.quantumpayments.io/post/agentic-commerce-payment-readiness-bottleneck

  • Meta title: Agentic Commerce Readiness: Why Payment Readiness Beats Protocol Bets

  • Meta description: Agents are closing holiday shortlists in October, but most merchants haven't made their payment stack agent-ready. Here is the readiness gap and the checklist that closes it.

  • Primary angle: Agentic commerce has moved into production and AI agents are finalising holiday shortlists in October. The urgent merchant problem is not choosing a winning protocol. It is closing the discovery-to-authorisation gap with machine-readable data, delegated intent controls, agent-aware fraud decisioning and defensible reconciliation.

  • Relevant tags: Agentic commerce, agentic payments, agent payments protocol, agentic checkout, Mastercard Agent Pay, payment orchestration, AI fraud detection payments, payment success rate, AI payments, unified commerce

  • Organisation tags: Quantum Payments, Fintech, Payments Technology, Payment Service Providers, Commerce Infrastructure

  • Hashtags: #AgenticCommerce #AgenticPayments #AIPayments #PaymentOrchestration #PaymentSecurity #PaymentSuccessRate #Fintech #QuantumPayments

Image CDN URLs

LinkedIn post 1, 8:08am AEST

Angle note for Sonny: Data-led strategic hook. Use the October timing to show that discovery is accelerating faster than authorisation. This is materially different from a generic “merchants are not ready” message: the point is that agents are narrowing holiday demand now, and payment readiness determines who converts it.

Visual concept: Use the hero image showing the AI shortlist funnel stopping before the payment authorisation gate.

Relevant tags: American Express, Shopify, Mastercard, EMVCo, Quantum Payments

Copy:

65% of shoppers plan to use AI for at least one shopping task this holiday season.

AI agents are already narrowing product shortlists in October.

But the readiness numbers expose the real bottleneck:

62% of merchants expect AI agents to have a moderate or significant impact within the next year.

Only 41% are experimenting with or implementing agentic commerce.

And while consumers are open to AI-led research and recommendations, only 15% are comfortable allowing an agent to buy high-value electronics.

That is the discovery-to-authorisation gap.

Agents can find the merchant, compare the product and recommend the purchase. The payment stack still needs to identify the agent, verify delegated intent, apply the right fraud decision and produce evidence when something goes wrong.

The first agentic holiday winners will not necessarily be the merchants with the loudest AI strategy.

They will be the merchants whose payment operations are ready when the shortlist becomes a transaction.

First comment:

The practical answer is not to bet the roadmap on one agent payments protocol. Keep product and policy data machine-readable, then use a PSP or orchestration layer to abstract the changing standards above a stable payment and reconciliation core.

LinkedIn post 2, 3:23pm AEST

Angle note for Sonny: Operator and merchant checklist. Focus on what payment, commerce and risk teams can implement immediately before agent traffic scales. Do not use the generic “merchants are not ready” framing. The angle is operational control before the October shortlist turns into holiday orders.

Visual concept: Use the inline image showing structured catalogue data separated from a fragmented payment stack, or the control-plane visual showing identity, intent, fraud and reconciliation signals converging.

Relevant tags: Mastercard, Google, EMVCo, Worldline, ING, Quantum Payments

Copy:

AI agents are moving from product discovery towards real payment execution.

Before agent traffic scales,run this merchant readiness checklist:

  1. Identify the agent and capture the relevant agent transaction signal.

  2. Store delegated intent limits including budget, category, supplier, recurrence and time window.

  3. Make returns and refund policies machine-readable.

  4. Preserve dispute evidence linking the instruction, delegation, product, price, authorisation and fulfilment record.

  5. Apply agent-aware fraud rules instead of blocking all machine-speed behaviour.

  6. Report payment success rate separately for assisted, delegated, human and unknown-origin traffic.

  7. Reconcile the full lifecycle from order and authorisation through capture, refund, chargeback and settlement.

  8. Map liability before launch: who is responsible when the agent acts outside scope or buys the wrong item?

The winning merchant stack will not depend on predicting which agent payments protocol wins.

It will depend on being able to operate across them without losing control of risk, evidence or reconciliation.

First comment:

The October shortlist is not only a marketing event. It is a payment operations test. If the catalogue is machine-readable but intent, fraud, refunds and reconciliation are not, the merchant may be discoverable without being buyable.

 
 
bottom of page