If agentic commerce is going to work at scale, the market has to solve more than authentication, aka the “identity problem,” as we posited in part two of this series exploring the six fault lines of agentic commerce.
When an AI agent enters a transaction flow, proving that a credentialed device, approved client, or linked account acted is not the same as proving that the human principal intended the specific transaction that resulted. That distinction matters because identity is doing more legal and commercial work than most systems admit.
But even if the market gets more comfortable with identity, the next problem is waiting right behind it. Once the system can say who appears to be acting, it still has to answer a second question that is just as important and often harder: What, exactly, was the AI agent authorized to do?
