Securing the Era of AI Agent Payments
By Ali Paterson · 19 November 2026
In brief: Mastercard is developing the trust framework for AI agents to conduct payments, focusing on verifiable intent and cross-rail interoperability to ensure secure, traceable transactions.
| Series | Spotlight on Mastercard |
|---|---|
| Key Speaker | Johan Gerber, Mastercard |
| Organization | Mastercard |
| Primary Market | Global |

Quick Summary
Johan Gerber of Mastercard joins this FF News Spotlight episode to discuss the evolution of agentic commerce. As AI agents begin to handle transactions on behalf of consumers, the industry must solve complex challenges regarding identity, authentication, and trust to ensure the safety of the global financial ecosystem.
Why is trust the foundation of agentic commerce?
Johan Gerber explains that for an agentic transaction to be successful, trust must follow the entire lifecycle of the payment. In a scenario where a consumer asks an agent to purchase a specific item, such as a red pair of sneakers, multiple layers of verification are required:
- The agent must verify it is dealing with a real consumer.
- The consumer must be certain they are interacting with a legitimate agent.
- The agent needs to ensure the merchant on the other end is a real business that will actually ship the goods once the payment is processed.
Without this end to end verification, the risk of fraud and fake businesses increases, making it difficult for autonomous commerce to scale.
What is the concept of verifiable intent?
To manage the risks associated with autonomous agents, Mastercard focuses on the concept of verifiable intent. Johan Gerber describes this as a process where the original instruction from the consumer is encrypted, signed, and stored securely. This creates a digital audit trail that can be accessed if a dispute arises. By opening this box, stakeholders can determine exactly what the agent was asked to do, whether the consumer provided the correct instruction, and if the agent executed that command accurately. This traceability is essential for determining liability and understanding who was at fault when something goes wrong in an automated transaction.
How will agents navigate different payment infrastructures?
One of the most significant shifts in this new era is how agents will interact with existing financial networks. Johan Gerber suggests that agents will eventually be capable of selecting the most appropriate infrastructure based on the specific needs of a transaction. For example, an agent might use a card rail to perform a risk assessment, an RTP rail for immediate money movement, or the SWIFT network for cross-border transfers. This interoperability allows for a more flexible and innovative payment ecosystem where the technology adapts to the specific requirements of the commerce use case.
FF NEWS TAKE:
The transition from AI as a search tool to AI as a transacting entity represents a fundamental shift in the global economy. By prioritising verifiable intent, Mastercard is addressing the potential black box problem of autonomous commerce before it becomes a systemic risk. The ability for agents to intelligently route payments across different rails suggests a future where the underlying infrastructure becomes invisible to the end user, leaving trust and security as the primary differentiators for financial institutions in the age of AI.
Frequently asked questions
What is agentic commerce?
Agentic commerce refers to a system where AI agents perform transactions and make purchases on behalf of human consumers, requiring new frameworks for identity and trust.
How does verifiable intent work in payments?
It involves encrypting and signing a consumer's instructions to an agent and storing them securely. This allows the instruction to be verified later to resolve disputes or errors.
Which payment networks will AI agents use?
AI agents may use a variety of infrastructures depending on the task, including card rails for risk assessment, real-time payment rails for execution, and SWIFT for cross-border movement.