Source: Nate’s Substack
As AI agents increasingly attempt transactions, companies are gatekeeping agent usage through existing operational infrastructure—fraud detection, pricing tiers, and onboarding flows—rather than building agent-specific policies. This means a company's anti-fraud team or finance department is effectively deciding whether Claude or a custom agent can become a customer, often without explicit intent or coordination. The competitive question is whether your current guardrails accidentally exclude agents or accidentally admit bad actors.