[comte7092]

How would chatgpt be able to spot overruns, fraud, or make a value assessment off of a budget? Unless you provided significantly more information than implied in your post, eg actual spend, the entire exercise was pretty absurd.

This is the biggest danger of LLMs: people assuming that they have some sort of magical super intelligence.

[Grimblewald]

Even if you did provide the data, if its tabular you can forget chatGPT understanding it properly, unless it is a very small table or without writing code to summarise things. If it writes code, theres a significant chance it still messes things up unless what you're asking is incredibly routine.