Scams are getting more and more sophisticated. We've always known that when you write a playbook, sooner or later the red team will find holes in that playbook. Perhaps that's where ML/AI comes in, since you can train them, but you never really understands what they really "learnt" from that training (/s sort-of, but the famous amazing snow/husky classifier always comes to mind)
Personally, I've received such calls before, and the first thing I'd ask for is a case number, and that I would call the support number printed on my credit card, to get back to them. Of course, if someone co-opts that number, then I'm also SOL, but I'd imagine then this would be engineering on a larger scale, rather than a specifically targeted whaling attempt.