The best AI voice generation is already much more realistic than the typical TikTok robotic voices, and they're only going to get better. It won't be long before the only reason anyone will generate artificial voices that are obviously fake will be choice, not tool limitations.
As genuine customer service migrates to robots, it may change expectations as to what sounds legitimate? So in the future the genuine human ringing you up would smell like a scam whereas the genuine robot will have weight - irregardless of its legitimacy?