When your AI agent needs to create an account on a new service, what happens?
TL;DR
A product manager at one of the largest Nordic payment companies is building AI agents on the side and runs into a concrete wall: agents handle existing accounts fine but break down completely when signing up for a new service.
Key Points
- Standard onboarding flows — forms, email OTPs, CAPTCHAs, ToS acceptance, and eventually KYC — force a human back into the loop every single time.
- He is considering building a dedicated service to solve this but wants to validate whether it is a widespread problem before investing the effort.
Nauti's Take
This is not an edge case — it is a systemic question. Anyone building agents that reach beyond a single workspace knows this moment: the agent stands in front of a signup form like a dog at a closed door.
'Human steps in briefly' does not scale by definition. Either specialized services emerge — a kind of passport-as-a-service for agents — or major platforms introduce machine-readable onboarding APIs.
Until then, humans remain the gatekeepers, and autonomous agents are only as autonomous as their weakest signup link.
Sources
