This is what invariably happens when you give bots control of important day-to-day business operations: The bot makes some horrendous mistake and then there's no human being around who has the authority, the access, and the knowledge to both revert the bot's decision and make sure that the bot doesn't just replay the error at a later date.

Privatizing gains, socializing loses

Bots and automation aren't inherently bad, but often times the motivation is pure cost savings for the company and users pay the price.

A more rational design could have a periodic human auditing process, appeal/reversal process, and a public audit/action log for clarity. However, that's going to eat into the savings of automating the whole thing so why bother