Programs, protocols and rules for AI agents that shop and pay on a person or business behalf, with the opportunity each one opens up.
Updated monthly · Last updated: · Items leave the tracker once their month has passed.
Who shapes agent payments
| Group | Programs and standards | Payments angle | Who it affects |
|---|---|---|---|
| Visa | Visa Intelligent Commerce and Intelligent Commerce Connect | Agent tokens, spend controls and authentication | Merchants, agent platforms, issuers |
| Mastercard | Agent Pay, agentic tokens and an agent payments protocol | Agent identity and verifiable consent | Issuers, merchants, agent platforms |
| Industry protocols | Google Agent Payments Protocol, OpenAI Agentic Commerce Protocol | Common ways to show intent and identity | Agent builders and merchants |
What changed this month
A short log of new and changed items, updated every month.
Week of Nov 16, 2026
- Tracker launched with 4 agent payment programs and protocols.
- Most items are rolling out in pilots; no fixed deadlines yet.
- AI rules for credit decisions stay on the Data Governance tracker, and AI risk guidance on the Cloud tracker.
AI and agent payments tracker
Soonest first. Tap a group or a status to filter; tap again to clear.
FILTER BY GROUP
FILTER BY STATUS
| Group | Rule or change | What it means | Opportunity | Key date |
|---|---|---|---|---|
| Visa | Visa Intelligent Commerce Connect rolls out Visa Our guide → · Source: Visa | One integration for agents to pay with tokens, spend controls and authentication; in pilot and expanding. | Agent-ready checkout and token controls for merchants and platforms. | Rolling out 2026 Planned |
| Mastercard | Agent Pay enabled for all issuers in Europe Mastercard Our guide → · Source: Mastercard | All European issuers are enabled at network level; live agent payments use passkeys. | Issuer controls and consent screens for agent purchases. | Live in pilots Planned |
| Mastercard | Protocol for AI agents to pay each other Mastercard with partners Our guide → · Source: Fortune | Records the permissions people give their AI agents so payments can be checked. | Machine-to-machine payment and permission checks for business flows. | Launched Jun 2026 Planned |
| Industry protocols | Agent payment protocols from Google and OpenAI Google AP2; OpenAI Agentic Commerce Protocol Source: Mastercard | Shared protocols for user intent, secure credentials and agent identity are forming. | Support more than one agent protocol in checkout and risk tools. | Evolving Planned |
WHO OWNS THIS
AI and agent payments: roles and sign-off
Each program or pilot needs a named owner. The agentic commerce product lead owns the pilot; the payments architect owns tokens and protocols; fraud and identity owns agent checks and limits; legal and compliance owns consent and liability; the head of digital signs off.
1. Roles and responsibilities
- Writes the impact assessment
- Chooses partners and protocols
- Runs go/no-go
- Reports results
- Designs agent token flows
- Compares protocols
- Reviews partner APIs
- Signs the design
- Sets spend limits and step-up checks
- Verifies agent identity
- Monitors fraud and disputes
- Certifies controls
- Builds token and checkout changes
- Logs agent actions
- Supports testing
- Fixes defects
- Writes consent and terms
- Assesses dispute liability
- Checks AI and privacy rules
- Certifies compliance
- Tracks network programs
- Signs partner agreements
- Coordinates pilots
- Confirms partners are ready
- Sets priorities and funding
- Accepts remaining risk
- Gives final sign-off
- Reports to leadership
2. Who is accountable for what (RACI)
| Activity | Product lead | Architect | Fraud and identity | Engineering | Legal and compliance | Partnerships | Sponsor |
|---|---|---|---|---|---|---|---|
| Log a network or protocol change | A | C | I | I | I | R | I |
| Impact assessment | A | R | C | C | C | C | I |
| Agent identity and consent design | C | C | A | C | C | I | I |
| Build and token integration | C | C | C | A | I | I | I |
| Liability, disputes and terms | C | I | C | I | A | C | I |
| Pilot go-live sign-off | R | R | R | R | R | R | A |
A accountable (one per activity, approves) · R responsible (does the work) · C consulted · I informed.
3. From review to sign-off
4. Sign-off record
One line per pilot or launch. No signed line, no go-live.
| Item | Prepared by | Reviewed by | Signed off by | When | Evidence |
|---|---|---|---|---|---|
| Visa Intelligent Commerce pilot | Agentic commerce product lead | Fraud and identity lead | Head of digital | Before pilot launch | Consent records, test logs |
| Mastercard Agent Pay (Europe) | Payments architect | Legal and compliance lead | Head of digital | Before pilot launch | Design, partner agreement |
Titles differ by firm; small teams may combine roles, but the person who builds should not be the only one who signs off. General guidance, not legal advice.
Runbook: from an agent payment program to sign-off
How to handle a new agent payment program, protocol or rule from this tracker. Use one runbook per item. Each phase says when, why, what, who and how, and the evidence to keep.
▸1. Plan · Decide if the item applies to you, how big the change is, and who owns it.
When: As soon as an item appears on the tracker. Target: at least 90 days before the key date (T-90). For proposed rules, inside the comment window.
Why: Late discovery is the main reason teams miss deadlines, pay fines or ship rushed changes. Early planning turns a rule into a normal roadmap item.
What: Decide if the item applies to you, how big the change is, and who owns it.
Who: Owner: Agentic commerce product lead. With: Partnerships lead (network and protocol news), Payments architect, Legal and compliance lead.
Example: a new agent payment program from a network. Confirm channels and customers in scope.
How
- Open the official source linked in the tracker and read the actual text, not only the summary.
- Map it to your products, rails, systems, vendors and customers in scope.
- Run a short gap analysis: what we do today versus what the rule needs.
- Size the work (small, medium, large) and add it to the roadmap with a target date.
- Log risks, assumptions, issues and dependencies in the RAID log.
Evidence to keep
- Impact assessment (applies / does not apply, with reason)
- Gap analysis
- Epic or change request with owner and target date
- RAID log entry
▸2. Prepare · Turn the rule into requirements, a design, a test plan and a rollout plan.
When: T-90 to T-60 days.
Why: Clear requirements and a test plan stop rework and give auditors a straight line from the rule to the change.
What: Turn the rule into requirements, a design, a test plan and a rollout plan.
Who: Owner: Agentic commerce product lead. With: Payments architect (token and protocol design), Fraud and identity lead (agent identity), Engineering lead.
Example: map consent, spend limits and agent identity to the design.
How
- Write requirements that quote the rule clause they come from.
- Agree the design and any vendor or partner changes, with dates.
- Build a traceability matrix: rule clause, requirement, test case, evidence.
- Plan test data, environments and a rollback.
- Book change windows and tell support and operations teams.
Evidence to keep
- Requirements with rule references
- Design or configuration decision
- Traceability matrix (draft)
- Test plan and rollback plan
▸3. Execute · Build or configure, test, and release the change safely.
When: T-60 days to go-live.
Why: This is where the change is built and proven. Testing against the rule, not just the feature, is what protects you.
What: Build or configure, test, and release the change safely.
Who: Owner: Engineering lead. With: Payments architect, Fraud and identity lead, network or wallet partner.
Example: connect agent tokens with spend controls.
How
- Build or configure the change behind a flag where possible.
- Run unit, system, regression and negative tests, plus UAT with the business.
- Fix defects and re-test; keep the defect log current.
- Hold a go/no-go meeting with clear criteria.
- Release, watch live metrics and keep the rollback ready.
Evidence to keep
- Test results and defect log
- UAT sign-off
- Go/no-go record
- Change ticket and release notes
▸4. Evidence · One evidence pack that shows the rule was understood, built, tested and approved.
When: Collected all the way through; packaged by T-7 days.
Why: If it is not written down, it did not happen. Good evidence makes audits short and calm.
What: One evidence pack that shows the rule was understood, built, tested and approved.
Who: Owner: Fraud and identity lead. With: Engineering lead (test logs), Legal and compliance lead (consent records).
Example: keep consent records, test logs and dispute samples.
How
- Create one folder per tracker item with a simple index.
- Store each document with date, version and owner.
- Complete the traceability matrix with links to every test and approval.
- Add screenshots or exports of live configuration where relevant.
- Record training or communications sent to staff and customers.
Evidence to keep
- Evidence index
- Final traceability matrix
- Configuration proof
- Training and communication records
▸5. Audit · Second line (Compliance or Risk) tests the control; Internal Audit may sample it later.
When: From T-7 days to 30 days after go-live.
Why: An independent check finds gaps before a regulator, scheme or customer does.
What: Second line (Compliance or Risk) tests the control; Internal Audit may sample it later.
Who: Owner: Legal and compliance lead (control check). Independent check: Internal audit.
Example: sample agent purchases for consent and limits.
How
- Walk through the evidence pack against the rule text.
- Re-perform a sample of key tests or checks.
- Raise findings with a rating, owner and due date.
- Track findings to closure and re-test the fix.
Evidence to keep
- Control test results
- Findings log with owners and dates
- Closure evidence for each finding
▸6. Review · Post-implementation review: did it work, what broke, what we learned.
When: About 30 days after go-live.
Why: A short look back confirms the change works in real life and makes the next rule easier.
What: Post-implementation review: did it work, what broke, what we learned.
Who: Owner: Agentic commerce product lead. With: Fraud and identity lead, Partnerships lead (pilot results and disputes).
Example: review fraud, disputes and conversion in the pilot.
How
- Check live metrics: errors, rejects, exceptions, complaints, incidents.
- Confirm all findings and defects are closed or accepted.
- Note lessons learned and update this runbook if needed.
- Update the tracker status and any related guides.
Evidence to keep
- Post-implementation review notes
- Metrics snapshot
- Lessons learned
▸7. Sign-off · Named people confirm the item is done and the evidence is complete.
When: After the review, and before closing the item.
Why: Formal sign-off shows clear ownership and closes the loop for auditors.
What: Named people confirm the item is done and the evidence is complete.
Who: Owner: Head of digital (approves). Signers in order: Agentic commerce product lead, Payments architect, Engineering lead, Fraud and identity lead, Legal and compliance lead, Partnerships lead.
Example: close the item when every signer has signed.
How
- Share the evidence pack and review notes with the approvers.
- Each approver confirms their part in writing, with date.
- Archive the pack in your GRC tool or records system.
- Close the epic and the RAID entries.
Evidence to keep
- Signed approvals (name, role, date)
- Archived evidence pack
- Closed epic and RAID items
Sign-off
Sign in this order: Prepared, then Reviewed, then Certified, then Approved. Each signer checks the evidence of the lines above. Close the item only when every line has a name, date and evidence link. The person who prepares cannot approve.
| # | Role | Stage | Confirms | Evidence | Name and date |
|---|---|---|---|---|---|
| 1 | Agentic commerce product lead | Prepared | Impact assessment done; pilot scope agreed | Impact assessment | ________________ |
| 2 | Payments architect | Reviewed | Token and protocol design correct | Design document | ________________ |
| 3 | Engineering lead | Reviewed | Integration built and stable | Change ticket | ________________ |
| 4 | Fraud and identity lead | Certified | Agent identity, limits and fraud checks work | Test results | ________________ |
| 5 | Legal and compliance lead | Certified | Consent, liability and terms covered | Legal memo | ________________ |
| 6 | Partnerships lead | Certified | Network and partner agreements in place | Agreements | ________________ |
| 7 | Head of digital | Approved | Accepts remaining risk and closes the item | Signed approval | ________________ |
Tip: copy this runbook into your own tracker or GRC tool, one copy per rule. Dates are relative to the key date shown in the tracker (T-90 means 90 days before).
Risk and control self-assessment (RCSA): AI and agent payments
The main risks for this area, the controls that reduce them, and the risk left after those controls. 4 risks are specific to this area and 6 are shared by every tracker (C1 to C6). Owners match the roles in the Who owns this section above. Reviewed every quarter and checked by the second line. Preventive controls lower likelihood; detective and corrective controls lower impact. Read more in our RCSA guide.
10 risks assessed for this area. Controls bring every high inherent risk down to medium or low.
How to read the heat map: each dot is a risk. It starts at its inherent rating (no controls) and moves to its residual rating (after controls). Score = likelihood x impact: 1 to 6 Low, 8 to 12 Medium, 15 to 25 High. Mid-blue dots are this area; navy dots are shared risks.
QUARTERLY RCSA CYCLE
RISK AND CONTROL REGISTER
| ID | Type | Risk and cause | Control | Owner | Inherent | Residual |
|---|---|---|---|---|---|---|
| AG-R1 | This area | An agent pays without valid consent Cause: Consent not recorded or not checked | Verifiable consent records and spend limits on every agent token Preventive and corrective · Continuous | Fraud and identity lead | High 15 L3 x I5 | Medium 8 L2 x I4 Control: Partly effective |
| AG-R2 | This area | A fake agent makes purchases Cause: Agent identity not verified | Agent identity checks and trusted-agent lists Preventive · Continuous | Fraud and identity lead | Medium 12 L3 x I4 | Medium 8 L2 x I4 Control: Partly effective |
| AG-R3 | This area | Disputes rise because liability is unclear Cause: Terms and network rules not mapped | Clear terms and dispute rules agreed before launch Preventive and corrective · Per pilot | Legal and compliance lead | Medium 12 L3 x I4 | Low 6 L2 x I3 Control: Effective |
| AG-R4 | This area | The chosen protocol loses support Cause: Market still choosing standards | Support more than one protocol; review quarterly Preventive and corrective · Quarterly | Payments architect | Medium 9 L3 x I3 | Low 4 L2 x I2 Control: Effective |
| C1 | Shared | A rule change is missed Cause: No owner watches the sources; news arrives late | Weekly tracker review; every new item gets a named owner within 5 days Preventive and detective · Weekly | Partnerships lead | High 20 L4 x I5 | Medium 8 L2 x I4 Control: Effective |
| C2 | Shared | A rule is read the wrong way Cause: Summaries used instead of the rule text; no expert review | The impact assessment is signed against the source text Preventive · Per item | Legal and compliance lead | High 15 L3 x I5 | Medium 10 L2 x I5 Control: Effective |
| C3 | Shared | A deadline is missed Cause: Work starts too late; dependencies not tracked | Planning starts at T-90; RAID log and go/no-go before release Preventive and detective · Per item | Agentic commerce product lead | High 16 L4 x I4 | Low 6 L2 x I3 Control: Effective |
| C4 | Shared | Testing does not cover the rule Cause: Tests check features, not rule clauses | Traceability from each rule clause to a test, with sign-off Preventive · Per release | Fraud and identity lead | Medium 12 L3 x I4 | Low 4 L1 x I4 Control: Effective |
| C5 | Shared | Evidence is missing at audit time Cause: Documents spread across tools; no index | Evidence pack with an index, checked at T-7 Preventive · Per item | Legal and compliance lead | Medium 9 L3 x I3 | Low 3 L1 x I3 Control: Effective |
| C6 | Shared | Audit findings stay open Cause: No due dates or follow-up | Findings log with owner and due date; monthly review Preventive and detective · Monthly | Legal and compliance lead | Medium 9 L3 x I3 | Low 4 L2 x I2 Control: Partly effective |
Open actions
- AG-R1: add consent logs to every agent purchase. Owner: Fraud and identity lead.
- AG-R4: review protocol support each quarter. Owner: Payments architect.
- Test two controls in this area each quarter and keep the results in the evidence pack. Owner: Legal and compliance lead.
Quarterly attestation
Agentic commerce product lead (first line): confirm risks, controls and ratings are current. ________________
Legal and compliance lead (second line review): confirms the review and challenge. ________________
Head of digital: accepts residual risk. ________________
This is a sample RCSA for planning and training. Adjust risks, ratings and owners to your own organisation before use. The full cross-area register is in the monthly portfolio status report.
This tracker is a summary for planning, not legal advice. Dates can change, so always check the official source before acting. The Opportunity column is our own view. Summaries are written in our own words; linked pages remain the property of their publishers. Corrections: sp@paymentconsultant.top.
© 2026 SP Sankar · paymentconsultant.top. This tracker, its summaries, diagrams and layout are original work and are protected. Please do not copy, republish or scrape this page without written permission. You are welcome to link to it. See also Payment Rules Tracker and Payment Tech News.