Does KYC change when an AI agent is the one making the payment?

I've been thinking about what compliance looks like once AI agents start actually spending money because the normal KYC model assumes there's a person or business behind the transaction so for example a user has already been verified and then gives an agent permission to make purchases for them. The agent might choose the merchant then decide when to buy and initiate the payment without the user approving every individual transaction. I'm not sure whether you should think of that as the verified user making the payment through software or whether the agent eventually needs some kind of identity and risk profile of its own. It gets even weirder if one person has several agents with different permissions. One might only be allowed to pay for software under $100 while another can make larger purchases from approved vendors. Treating both as normal activity from the same user seems like you're throwing away useful context about who or what actually initiated each transaction. My instinct is that you still KYC the human or business behind the funds but then separately track what authority each agent has and monitor what it actually does with that authority. I wanna know how people building agent payments are thinking about this and like does KYC basically stay attached to the human/business with agents treated as delegated actors or do agents eventually need their own compliance identity too?