I'm thinking about doing a Phd in AI. I'm a software engineer.

Anyone done similar have advice?

As my advisor puts it, "only consider PhD if there is something that *you* want to do with it (some idea you want to explore or a field you want to master etc., this is the best case) and/or if it's necessary for a career that you want to be in (academics, applied academics, industrial research etc.)."

It really helps if you have a clear idea of what you want to do, why you want to do it, what are your expectations, and what the opportunity cost is. It's like half a decade of disorganized (or less organized anyway) work with fewer tangible benefits (eg. you usually get paid much less than industry average). Of course there are benefits too, but whether the trade-off is acceptable to you depends on you.

Then of course, there's getting into a PhD program. Do you have an undergrad/masters degree? Relevant work experience? Research experience and/or publications?

And finally, the kind of thing you like to do, whether it fits what a PhD program offers. Research is a really weird environment, you need to enjoy it or it's going to be a pretty miserable half-decade. It's much less organized, more self-driven, and knowledge (and funding (and advisor)) centric. Many engineering PhDs at least have industry connections, so you do work on relevant stuff, but it's a lot more self-driven than a job in the same industry.

This is from my experience, others (especially from other countries) might vary.