Fifteen hours of recordings, and no idea who said what

I did an experiment last year which turned out to be a disaster. I was recording the full day. When I entered the hall, I turned on my mic, and when I went out of it, I turned it off. So I recorded every chat in a hall, every session. The challenge I had after three days, which means fifteen or whatever hours of recordings, I had zero chance of figuring out who said what. What was a presentation? What was a random chat in a hall? Who is who? So I need to do a better job this year. I plan to record everything in an ideally more structured way, and AI is now better, so AI should help me do a better job with processing that.