[rokbenko]

Thanks for the question! That's a tough one. I still didn't figure out how to do it programmatically.

That said, keep in mind that I'm not using just an LLM. I equip the LLM with a tool to give it up-to-date knowledge from the web, which significantly reduces AI hallucinations. Still, it's not perfect.

What I've done so far is:

1) I tested the report for myself. It was 100% on point. 2) I tested the report with a few potential customers. None reported information in reports being incorrect.