The best AI agent in the world can't be fired. That's exactly why nobody will trust it with real money.

AI agents can be useful inside a strategy. But serious capital won't trust them without a human manager anytime soon. Performance is not the reason.
Investors don't just buy returns. They buy a person. Someone who owns the decisions, explains them when things go sideways, and has skin in the game. Reputation, career, sometimes a court case. That's what keeps a portfolio manager honest. Not the model. The stakes.
AI has none of that. And this is the one way AI agents can't be better than humans. They have nothing to lose — so nobody trusts them with the money.
When a strategy blows up, a PM gets on the call and walks the client through what happened. An agent produces logs. Logs are not accountability.
People say quant funds already run money without human judgment. They don't. Behind every systematic strategy sits a PM who built the model, defends it in due diligence, and gets fired when it stops working. The algo is a tool. The human decides.
Agents can work the same way under supervision, as a better version of existing algorithmic strategies. But then the human is still running the money. The agent is infrastructure, not the decision-maker.
I keep coming back to a simple test. Would you wire $50M to a system that can't answer for its mistakes and explains itself through token probabilities?
Current AI agents, especially the ones built around prompts, aren't going to replace the human in the loop. Nobody is responsible when they break. And without that, you don't have asset management. You have a slot machine.