People like to criticize ML, but my worst experience with Facebook was caused by a truly fucked up rule based system that banned my startup's Instagram account

People like to criticize ML, but my worst experience with Facebook was caused by a truly fucked up rule based system that banned my startup's Instagram account because the account operator changed the email address too quickly after merging it with a Facebook ads account. No machine learning, no probabilities, just "if (event that is usually sketchy) { ban with no appeal }" Even as a Facebook employee, there was nothing I could do to help (I joined Facebook after leaving this startup, but still wanted to help because I founded it). Even determining the cause of the ban could have cost me my job (for good reason, don't want corruption and bribery, etc). I understand the need to be cautious, but based on my analysis at the time the false positive rate was over 5% for this particular rule (based on later reactivations). Seemed way to high considering my startup had spent over 50K on FB ads in the connected account.