When I advise someone looking to break into AI roles, I start with the kind of role they want, because that matters a lot to the advice.
I suggest learning ML models even to someone who does not want to become a data scientist.
I see roles slowly overlapping as AI assistants become available; even a simple data analyst uses ML models now.
That overlap is why I recommend learning ML models across role types, rather than treating them as relevant only to data scientists.