Agentic commerce turns trust from a one-time authentication event into continuous policy. The system must know which agent is acting, for whom, under what mandate, with what limits, using which payment credential, and whether the action is auditable and reversible. Visa and Mastercard are converging on cryptographic agent recognition, verifiable intent, consent and traceability, while consumer research shows willingness to delegate remains far behind technical capability. Identity, authorisation and policy enforcement—not model intelligence—are becoming the binding layer for autonomous economic action.
Research method
This report triangulates current primary/public-sector evidence, large-sample research and relevant industry/platform documentation. It prioritises observed behaviour over forecasts, distinguishes vendor claims from neutral statistics, and uses an explicit 8i judgment layer only where source evidence ends. The analysis is current through 15 September 2026.
1. Capability is ahead of trust
Visa’s September 2026 US Trust Index found 72% of respondents had used an AI assistant, but only 23% trusted generative AI to handle payments on their behalf. Trust rose to 61% when Visa was positioned as securing the transaction. S1 The exact brand effect should not be generalised beyond the survey, but the trust gap is unmistakable.
Mastercard’s July 2026 ‘Encoding Trust’ work reaches a similar conclusion: technical barriers are falling faster than willingness to delegate, making identity, intent, consent and control central to scale. S2
2. Delegation creates a new identity graph
Traditional commerce usually authenticates a person/account and authorises a transaction. Agentic commerce introduces at least four identities: the human or organisation, the agent, the merchant/service and the payment/credential context. The system also needs a mandate linking them.
A robust architecture must answer: Is this a recognised agent? Who controls it? Is the user authenticated? What exactly was delegated? Are amount/category/time limits valid? Has the instruction changed? Can the action be attributed later?
3. Industry protocols are converging on similar primitives
| Primitive | Visa direction | Mastercard direction | Why it matters |
|---|---|---|---|
| Agent identity | Trusted Agent Protocol uses cryptographic signatures and recognised agents. | Know Your Agent registration and traceable identifiers. | Separates legitimate delegated agents from bots/impersonators. |
| User mandate | Signed/verified intent carried through the interaction. | Verified order intent and explicit consent. | Prevents an agent from inferring authority. |
| Credential security | Trusted commerce/payment framework. | Agentic/network tokens and passkeys. | Limits credential exposure and supports traceability. |
| Policy | Merchant verification and transaction context. | Permissioned rules, limits and continuous governance. | Turns authorisation into enforceable delegation. |
| Auditability | Signature validation and merchant visibility. | Traceable agent transactions. | Supports disputes, liability and compliance. |
4. Trusted Agent Protocol: recognition before transaction
Visa’s Trusted Agent Protocol provides merchants with a cryptographic method to distinguish approved commerce agents from anonymous bots, including message-signature validation and intent signalling. S3S4 This solves an immediate operational problem: bot-mitigation systems historically treat automation as hostile, while future merchants need to recognise permissioned agents as a legitimate customer channel.
Sources and notes
- Visa, Trust Index for agentic commerce (9 Sep 2026). [Source](https://usa.visa.com/about-visa/newsroom/press-releases.releaseId.22736.html)) — Harris Poll of 2,065 US consumers.
- Mastercard Signals, Encoding Trust (Jul 2026). [Source](https://innovationinsights.mastercard.com/encoding-trust)) — Industry perspective on identity, intent, consent and control.
- Visa Developer, Trusted Agent Protocol — Getting Started. [Source](https://developer.visa.com/capabilities/trusted-agent-protocol/docs)) — Technical merchant/agent recognition guidance.
- Visa Developer, Trusted Agent Protocol specifications. [Source](https://developer.visa.com/capabilities/trusted-agent-protocol/trusted-agent-protocol-specifications)) — Protocol scope and merchant mechanisms.
8i TECHNOLOGIES
Trust, Identity & Delegated Decision-Making
The Control Plane for Agents Acting, Buying and Paying on Behalf of People and Organisations
Reference date: 15 September 2026
See what matters next.
| Executive answer Agentic commerce turns trust from a one-time authentication event into continuous policy. The system must know which agent is acting, for whom, under what mandate, with what limits, using which payment credential, and whether the action is auditable and reversible. Visa and Mastercard are converging on cryptographic agent recognition, verifiable intent, consent and traceability, while consumer research shows willingness to delegate remains far behind technical capability. Identity, authorisation and policy enforcement—not model intelligence—are becoming the binding layer for autonomous economic action. |
|---|
Research method
This report triangulates current primary/public-sector evidence, large-sample research and relevant industry/platform documentation. It prioritises observed behaviour over forecasts, distinguishes vendor claims from neutral statistics, and uses an explicit 8i judgment layer only where source evidence ends. The analysis is current through 15 September 2026.
1. Capability is ahead of trust
Visa’s September 2026 US Trust Index found 72% of respondents had used an AI assistant, but only 23% trusted generative AI to handle payments on their behalf. Trust rose to 61% when Visa was positioned as securing the transaction. [S1] The exact brand effect should not be generalised beyond the survey, but the trust gap is unmistakable.
Mastercard’s July 2026 ‘Encoding Trust’ work reaches a similar conclusion: technical barriers are falling faster than willingness to delegate, making identity, intent, consent and control central to scale. [S2]
2. Delegation creates a new identity graph
Traditional commerce usually authenticates a person/account and authorises a transaction. Agentic commerce introduces at least four identities: the human or organisation, the agent, the merchant/service and the payment/credential context. The system also needs a mandate linking them.
A robust architecture must answer: Is this a recognised agent? Who controls it? Is the user authenticated? What exactly was delegated? Are amount/category/time limits valid? Has the instruction changed? Can the action be attributed later?
3. Industry protocols are converging on similar primitives
| Primitive | Visa direction | Mastercard direction | Why it matters |
|---|---|---|---|
| Agent identity | Trusted Agent Protocol uses cryptographic signatures and recognised agents. | Know Your Agent registration and traceable identifiers. | Separates legitimate delegated agents from bots/impersonators. |
| User mandate | Signed/verified intent carried through the interaction. | Verified order intent and explicit consent. | Prevents an agent from inferring authority. |
| Credential security | Trusted commerce/payment framework. | Agentic/network tokens and passkeys. | Limits credential exposure and supports traceability. |
| Policy | Merchant verification and transaction context. | Permissioned rules, limits and continuous governance. | Turns authorisation into enforceable delegation. |
| Auditability | Signature validation and merchant visibility. | Traceable agent transactions. | Supports disputes, liability and compliance. |
4. Trusted Agent Protocol: recognition before transaction
Visa’s Trusted Agent Protocol provides merchants with a cryptographic method to distinguish approved commerce agents from anonymous bots, including message-signature validation and intent signalling. [S3][S4] This solves an immediate operational problem: bot-mitigation systems historically treat automation as hostile, while future merchants need to recognise permissioned agents as a legitimate customer channel.
5. Mastercard: Know Your Agent and machine payments
Mastercard Agent Pay uses a ‘Know Your Agent’ trust model in which registered agents are identifiable and traceable, with user intent and consent bound into payment flows. [S5] Its June 2026 Agent Pay for Machines initiative extends the concept to machine-to-machine microtransactions and programmatic services. [S6]
The strategic point is broader than payments: as agents become economic actors, policy-controlled identity becomes part of the application architecture.
6. Delegation should be granular, not binary
- Category: what types of good/service may be purchased or action taken?
- Amount: per-action, daily and aggregate limits.
- Time: one-off, recurring or expiring authority.
- Counterparty: approved merchants/services or open market.
- Data: which personal, enterprise or preference data may be disclosed?
- Negotiation: may the agent change price/terms or only accept within bounds?
- Confirmation: which actions require step-up human approval?
- Revocation: how quickly can authority be withdrawn everywhere?
7. Liability and dispute become architecture problems
When an agent acts within an ambiguous instruction, the post-hoc question ‘who is responsible?’ becomes difficult. The better design is to make authority machine-verifiable before execution and keep a record of intent, agent identity, model/service version, policy checks and transaction result.
This changes audit from logging after the event to evidence encoded into the event.
8. 8i assessment
The durable layer in agentic commerce may not be the shopping agent itself; it may be the trust infrastructure that makes delegated action acceptable across merchants and jurisdictions. 8i should monitor interoperability, agent identity standards, consumer delegation thresholds, live transaction volume and dispute/fraud outcomes as a distinct intelligence domain.
Source register
Primary and high-value sources used. Time-sensitive claims were checked against the latest available release as of the reference date. Commercial/vendor research is identified as such and is not treated as neutral market truth without corroboration.
[S1] Visa, Trust Index for agentic commerce (9 Sep 2026). Source — Harris Poll of 2,065 US consumers.
[S2] Mastercard Signals, Encoding Trust (Jul 2026). Source — Industry perspective on identity, intent, consent and control.
[S3] Visa Developer, Trusted Agent Protocol — Getting Started. Source — Technical merchant/agent recognition guidance.
[S4] Visa Developer, Trusted Agent Protocol specifications. Source — Protocol scope and merchant mechanisms.
[S5] Mastercard, Agent Pay. Source — Know Your Agent, verifiable intent and consent architecture.
[S6] Mastercard, Agent Pay for Machines (10 Jun 2026). Source — Machine-to-machine permissioned payments and partner ecosystem.
[S7] Mastercard, first agentic transactions in Moldova (8 Jul 2026). Source — Live pilot; authenticated/traceable transaction model.