I imagine a scenario where upholding the AI bot's decision is one click of a button for a human. But going against the bot's decision requires filling out an on

I imagine a scenario where upholding the AI bot's decision is one click of a button for a human. But going against the bot's decision requires filling out an onerous form and making an argument as to why the bot is wrong. And oh yeah, you're under quota to process so many appeals per hour. All so some VP's slide can tout how much money their AI-powered fraud detection system has saved in human labor costs and its 97% success rate.