There are two very different things people mean by “AI trading,” and the gap between them is the difference between a useful tool and a way to lose money confidently.
The hype version
The version that gets attention is AI that finds you trades, point a model at the market and it discovers profitable strategies. It’s seductive and it’s mostly a trap. Markets are adversarial and non-stationary; a model that “found” an edge in historical data has, far more often than not, found a pattern that won’t repeat. Worse, a black box that hands you trades gives you nothing to inspect, no reason to trust it, and no way to tell skill from a lucky backtest. Any tool that promises this should raise your guard, not lower it.
The useful version
The version that actually helps is quieter: AI as an assistant for building and debugging the system. Drafting strategy code from a description you refine. Reviewing logic for the off-by-one in a lookback or the sign error in sizing. Explaining why a run failed, given the real error, the code, and the logs. This is grounded, checkable work, you can read what it suggests and decide whether it’s right.
Notice the difference. The hype version asks you to trust an output you can’t verify. The useful version speeds up work you’d do anyway and still fully control.
Why the distinction matters
Because they point at opposite relationships with the technology. “AI finds trades” wants you to hand over judgment. “AI helps you build” keeps judgment with you and just makes you faster. The first is where credibility goes to die; the second is genuinely valuable and not very glamorous.
It also matters because the hard part of systematic trading was never generating ideas, it’s the infrastructure and operations around them: execution, fills, broker state, monitoring, diagnostics. AI doesn’t make that infrastructure unnecessary. At best it helps you build and debug it faster. The system still has to exist.
The honest position
Be enthusiastic about AI as a development assistant: drafting, reviewing, debugging, with a human approving every change. Be deeply skeptical of AI as an oracle that produces trades. The first is a real productivity gain. The second is a story.
The infrastructure still has to be built, the strategy is still your responsibility, and the most an honest AI tool can claim is that it helped you do both a little faster, not that it did your thinking for you.