AI agents payments governance requires clearer decision rights, Maverick says

AI agents payments governance is becoming a board-level operating issue as software moves from assisting staff to making and carrying out decisions inside payments businesses, Maverick Payments President and Chief Operating Officer Ben Griefer wrote in a PYMNTS eBook. Griefer said the commercial stakes include customer service, fraud outcomes, risk pricing and the ability to grow without adding proportionate exposure.

In the PYMNTS eBook, Building the Agent-Ready Payments Enterprise, Griefer drew a distinction between AI copilots that support employees and agents that can plan, choose a course of action and act across enterprise systems. That shift, he argued, forces payments companies to decide in advance which calls machines may make and which calls remain accountable to people.

Who is accountable when AI agents make payments decisions?

Griefer’s answer is that accountability stays with the company and its people, even when a third-party model or partner platform is involved. Customers, he wrote, expect an accountable party for an outcome and are unlikely to treat an AI system’s role as a substitute for ownership by the business.

The practical boundary is clearest in fraud management. According to Griefer, an agent that identifies a transaction as suspicious and sends it to a reviewer can improve speed without making the final commercial judgment. An agent that rejects a real customer or allows a fraudulent transaction through has made a business decision at the point where the firm may preserve or lose a customer relationship.

The same issue appears in chargebacks and reconciliation. Griefer said AI agents can gather evidence, compare transaction records and find mismatches more quickly than manual processes. He argued that final decisions on disputed high-value chargebacks or unresolved gaps with a partner bank should still require a person’s approval because errors could be multiplied across large transaction volumes.

Agentic AI in payments refers to systems that do more than generate recommendations. These tools can execute tasks across workflows, such as routing a case, compiling documentation or selecting data for review, depending on the permissions a company gives them.

Griefer said the business case is not confined to back-office efficiency. Agentic systems can support faster merchant onboarding, quicker dispute handling and shorter underwriting cycles. At Maverick Payments, he wrote, AI helps speed approvals, bring processor and portfolio data to teams and improve decision-making as the company connects merchants with suitable payments solutions.

His central governance point is that speed should be paired with defined decision rights. Payments firms, he wrote, should classify workflows by risk: those that can be automated end to end, those that need a human checkpoint and those that should remain outside full automation even if model performance improves.

Griefer also urged firms to be clear about where outside models and partner platforms sit in a workflow. In his view, licensing technology does not move responsibility away from the payments company using it.

The firms best placed to benefit from agentic payments, Griefer wrote, will be those that design governance before deployment rather than adding controls after systems are already operating. The measure of readiness, in his assessment, is not the number of agents a firm uses, but whether it can explain which decisions those agents are permitted to make and why.

This story draws on original reporting from PYMNTS.