That’s a really good question. The thing with AI is, it’s only as smart as the data you give it. If you only feed it the same type of scenarios, it’ll keep reinforcing the same patterns and completely miss the odd cases. What I’ve found helpful is to mix things up: give it a wide range of data, including those “weird” outlier situations that don’t happen often but are super important.
And honestly, I never rely on AI alone. It’s great at spotting patterns fast, but it doesn’t have the intuition we humans bring. So I use AI as a partner, it helps me cover a lot of ground, and I use my own tester’s instinct to catch the blind spots it can’t see.