Sometimes those systems are the absolute worst. haha.

I have at least two fairly epic Amazon battle stories that are similar, some of the most frustrating exchanges of my life. I'd actually go a step beyond what you said, in the case you describe - I think they deliberately design the system to get people to give up their complaints, rather than to deal with it. Just exactly the same way that companies make it very easy for happy customers to leave reviews, but then actively try to discourage unhappy customers from doing so.

I think the role of consumer protection laws, among other things, is to prevent companies from deciding that "customer satisfaction" isn't the metric they're training against, but "call resolutions" whether the customer is happy or not. It's to keep automated systems from being designed to be impossible to navigate deliberately to suppress certain types of interactions.

Like, as you mentioned, requiring a valid tracking numbers to talk to someone, when they have your tracking number and can look you up by name, is deliberately designed to discourage moderately annoyed complainers. In other words, the company can "save" a lot of time if you just make sure that only the people annoyed enough to make a big fuss about it are the ones who get through. Everyone else will just sort of burn themselves out, so stops being the company's problem.

I actually think that's exactly what consumer protection laws are for.

I think always having a option to talk to a human is a great thing. I'm just not sure it's possible to always require a company to be liable if the human opt-out is slower or less capable than the AI system. And doing so will discourage them from having better systems in place.

Right now, what you describe is the computer sucking, and the human being able to help. If there ever is a time when the AI is better and faster at resolving the issue, and the human worse, then the issue I'm seeing may become more realistic, I think.