Have you tried this and gotten anything useful? I do a lot of building with OpenAI Assistants/Agents and I'm not sure you'll get great results with this approach (though obviously it depends on the depth of your prompt and what kind of results you are looking to get).
IMO, what you are going to get back will be very surface level and not so interesting. In theory, you could supply data to ChatGPT and have it calculate Win-Rate, but it won't tell you WHY the win-rate is going up or down or how to fix it. There would be so much more data/steps/context needed for truly useful insights.
I think the best approach is to use AI to cut down on the time it takes to do the analysis, which should open your time up to think about the talking points. But, again, this is just my experience and maybe you or someone else has a better approach then I've tried.