I'd bet there's some calculation that people who try to sign up for a plan over the phone end up using the phone more down the line, which would mean more costly operator time. So the math works out where the overall savings of making enough people give up before reaching a human outweighs the cost of potentially lost new subscriptions by phone call. Or, they just didn't study that. Or, the decision-makers don't contact customer support for themselves and so don't know how infuriatingly unhelpful AI ones are.
Or the decision maker put "replaced 50% of call workers with AI" on their resume and got a new job instead of measuring the results.
This is not to say that we should not actively work to prevent criminal acts, but that trying to establish a world in which such acts are impossible will cripple society in ways which will leave us vulnerable to much larger and more systemic abuses. Benjamin Franklin’s statement rings as true as ever, if in a rather updated context: “ They who can give up essential liberty to obtain a little temporary safety, deserve neither liberty nor safety.”
And what do we do for the children who have parents who fail them. How do we even detect it in time to help those children?