I don't trust GPT to do multiple tasks at once.

I work with company research, so I use GPT a lot to find things like revenue, legal name, address, phone, executives, parent companies, subsidiaries, etc What I've noticed is that the more things I ask in the same prompt, the less I trust the result. For example, if I ask it to find the basic company info, identify the main executives and estimate the revenue all at once, usually one part feels much less researched than the others. Maybe the executives are right but the revenue estimate is weak, or it gets the company info right but completely misses an acquisition or parent company relationship. Because of that, I ended up using separate chats for basically everything. One for revenue, another for executives, another for company info, and so on. Is this an actual limitation with LLMs, or am I just being overly cautious? It feels a bit stupid because technically GPT should be able to handle all of this in one request, but in practice I get much more consistent results when I give it one specific job.