Comment by Greamy

It can benefit from more complex algorithms, but I would stay away from "AI" as much as possible unless there is indeed need of it. You can analyse your data and make some dynamic thresholds, you can make some small ML models, even some tiny DL models, and I would try the options in this order. Some cases do need more complex techniques, but more often than not, you can solve most of your problems by preprocessing your data. I've seen too many solutions where a tiny algorithm could do exactly what a junior implemented using a giant model that takes forever to run.