Agentic Payments & Commerce
When an AI agent pays, who is in control — and who pays when it goes wrong? At City Agentic Finance Lab (CAFiL), we research the mandates, protocols, rails and market design of agentic commerce.
What we explore
Agents that search, negotiate and pay; how people and firms delegate and revoke spending authority; liability and redress when agent-initiated payments go wrong, including scams; agent-to-agent markets; emerging protocols such as Google’s Agent Payments Protocol (AP2), the Agentic Commerce Protocol (ACP) from OpenAI and Stripe, x402, Mastercard Agent Pay and Visa Intelligent Commerce; merchant economics; and European payment sovereignty. We combine agent-based simulation, a field our members have worked in since 2003, with evidence from industry.
Questions we are asking
- How should people delegate — and revoke — spending authority to an agent?
- Who is liable when an agent-initiated payment goes wrong, and how should redress work?
- Which rails and protocols suit agent-initiated payments, and how do they compare?
- How will agentic commerce change merchant economics and competition?
- How can the UK and Europe avoid dependence on a few agent platforms?
What we produce
- Briefing paper (in preparation): “Who pays when the agent pays? Mandates, liability and redress in agent-initiated payments”
- Agent Sandbox experiments on agent-to-agent markets (“Bots bargaining”)
- Roundtables on payment mandates
- Working papers on agentic payments in Europe (in progress)