AI Agents Get Permission to Spend: Agentic Payments Take Center Stage

Agentic payments are emerging as a new way for AI agents to make transactions on behalf of users. This concept is already being tested in production environments, with companies like Santander and Mastercard completing live end-to-end payments using AI agents.

The key idea behind agentic payments is delegated authority, where the user gives the AI agent limited access to their funds, rather than unlimited control over their bank account or card. This can include setting spending limits, approved merchants, permitted categories, geographic or time limits, and extra approval above a certain amount.

Cards and stablecoins can both power agentic payments, with Visa's Machine Payments Protocol allowing agents to receive payment requests, authorize them programmatically, and settle using cards, stablecoins, or other supported methods. Mastercard's Agent Pay for Machines also supports settlement across cards, accounts, and stablecoins.

The hardest problem facing agentic commerce is trust, who is liable when an AI agent makes a mistake? Payment companies are focusing on identity, authorization, and auditability to address this issue, with India's proposed UPI framework exploring similar controls.