Meh .. staff with PhDs are good for different things than those with Masters. If you are applying established methodology, you don't need PhDs on your team. For new things or things for which there is no established methods, PhDs are essential IMHO.
Data science seems like it has a feedback issue. If you design an experiment poorly, it's easy to never know you learned the wrong thing. If you're building an ML model and it's less accurate that what an expert would build, you might never know either. When you build many other kinds of products, it's easier to know whether they're working. Data products don't seem that way. I wonder if that could lead to weird "market for lemons" sorts of things.
Well, if you don't understand what skills you're looking for other than the one buzzword "AI", then you almost certainly don't know how to measure those people. So then it becomes a credential / status game ("I have a Ph.D", "I went to Stanford", "I want twice as much money because I think I'm twice as good as anyone else", or whatever ideas your recruiters/execs have in their head is "good").
If you can't recognize the skills, you certainly can't recognize talent. Recognizing talent without using skills or credentials is still very hard for companies. Basically you need someone with the skills, credentials, or talent to recognize someone else worth having.
That being said, HR or even recruiters honestly don't do any evaluation of candidates anywhere I've ever been. It's all delegated to engineers and technical people. HR is just there to fill out paperwork, handle logistics, and cover their butts on legal matters. Should they be fired? I'll leave that up to you, but I think the skills of most HR people and teams are highly overrated, although necessary at some level.