[D] What are your horror stories from being tasked impossible ML problems

ML is very good at solving a niche set of problems, but most of the technical nuances are lost on tech bros and managers. What are some problems you have been told to solve which would be impossible (no data, useless data, unrealistic expectations) or a misapplication of ML (can you have this LLM do all of out accounting).

data good horror lost machinelearning managers set solve stories tech tech bros technical