Scrolling through my feed, there are two dominant ways I see traders actually using AI right now and neither of them is particularly flattering.
Dashboarding
There's something incredibly fulfilling about creating a system that's beautifully designed, tracking all the metrics anyone could ever want, P&L overlays, heat maps, sentiment feeds, whatever. It feels like real work. But for many people, it's just procrastination in a more expensive suit. The dashboard becomes the real project, and actual trading gets pushed to the side.
Backtesting/Overfitting
AI has made backtesting so accessible and so fast that people can torture historical data until it confesses to almost anything. And they do. I've literally watched people backtest a strategy, find that it underperforms on certain days of the week or certain hours, and simply remove those periods from the sample.
The irony is that AI was supposed to remove human error from the process, but in practice it's often just amplifying the same old biases with a lot more computational power and a lot more speed behind them. Garbage in, garbage out; except now the garbage is better formatted and executes faster.