Does agent-initiated commerce break existing payment infrastructure?

Katharina Schumacher and Roger Roberts, with Katharina Giebel, writing for McKinsey & Company (QuantumBlack), argue that the shift from direct human purchases to agent-initiated transactions breaks an assumption baked into existing payments infrastructure. In their words: "As commerce transitions from direct user interactions to agent-initiated transactions, the existing payments infrastructure will encounter significant structural challenges. Traditionally, components such as gateways and fraud engines have been built around a human-in-the-loop model, in which identity, intent, and authorization are explicit and observable." When an agent — not a human — clicks "buy," the signals that gateways and fraud engines were designed to read (a human physically present, typing a password, observably forming intent) no longer hold.

McKinsey's proposed mechanism for closing this gap is Google's Agent Payments Protocol (AP2), described as "a groundbreaking open, payment-agnostic protocol designed to empower autonomous and semiautonomous AI agents to make verifiable purchases on behalf of users." AP2 works, per the excerpt, "by utilizing cryptographically signed mandates that link intent, cart, and payment across users, merchants, and payment networks," which McKinsey says "creates an audit trail, not subject to repudiation, ensuring transparency and accountability." The claimed payoff is threefold: new revenue from automated "standing intents," lower fraud and chargeback costs, and a reshaping of competition as "agent routing algorithms optimize for cost, speed, and rewards." The signed mandate is offered as a replacement for the human-observable signals that fraud engines used to rely on.

This sharpens a problem that identity-infrastructure notes elsewhere treat more abstractly. Can existing identity standards work for AI agents? proposes reusing OAuth, OIDC, and SPIFFE/SPIRE rather than building new agent-specific protocols; McKinsey's account of AP2 is a counterexample in practice — a purpose-built agent protocol (signed mandates tying intent to payment) rather than an adaptation of existing enterprise identity standards, which raises the question of whether payments specifically need bespoke protocols even if other agent-identity problems do not. It also gives a concrete commercial setting for the authorization and proportionality gap named in Why do agents fail at identity verification and authorization?: the "human-in-the-loop" assumption McKinsey describes breaking is exactly the explicit-authorization signal that red-teaming found agents can fail to respect.

The excerpt does not establish whether AP2 has been adopted by any payment network, merchant, or card issuer at the time of writing, nor whether cryptographic mandates actually reduce fraud and chargebacks in practice rather than merely shifting liability or creating new attack surfaces (a compromised or manipulated agent could still sign a mandate the user never intended). It is McKinsey's characterization of a vendor protocol's design goals, not an independent or empirical evaluation of AP2 in use — the implication that this ensures transparency and accountability holds only to the extent the protocol is implemented correctly and widely adopted across the payments ecosystem it depends on.