Using AI up and through LOI?

searcher profile

December 19, 2025

by a searcher in Idaho Falls, ID, USA

A potential deal came my way a few weeks ago that aligns almost perfectly with my deal box. The seller does not and is not, interested in working with a broker. So far, only their CPA is involved. They are retiring but want a strong successor and will stay on for 6 mo in an advisory role. We have an NDA, high level financials and business structure. We have a long runway to close and I'm confident enough now to put in an LOI. Up to this point, I've created a master AI folder for the aquisition and inside of that folder, multiple LLM's that focus specificially on different parts of the process but all collaborate. I have one that think's like my copilot, one for negotiations, a CPA, an SBA broker, and so on. Any time I need more information or need to dive deeper into something I'm not familiary with, I ask there. I also drop every corespondence and every detail I can so it tracks timelines and learns as we go. It's been extremely helpful to date and I wouldn't be where I am in this process without it. My question: How far down the AI rabbit hole can I really go? It's been so good so far that I feel like I could run this entire deal through. But.... I know I need a real lawyer, cpa, etc at some point. Has anybody else done this successfully and how far did you go?
0
2
20
Replies
2
commentor profile
Reply by a searcher
in Orlando, FL, USA
generally, _verify verify verify_ - everything the LLM says could be made up, or accurate, or someone elses opinion. your team (it's time to ID them!) may or may not accept the drafts that you create with LLM and work from them. it's worth asking. that said - even as an exercise in seeing what a contract and terms _looks like_ it's a great way to learn and see what options you might have, and what to look for. then you're better armed when you DO pay by the hour....you're not asking "what does this mean?" because you've already seen it and asked. but use what you get to do your own research to validate it - you can't trust it on it's own. :)
commentor profile
Reply by a searcher
from University of Pennsylvania in Tucson, AZ, USA
I use and help build specific AI tools for a wide variety of things for my day job. Most of the time, I love it. Even with the latest models though, I have gotten some WILDLY wrong answers, especially when reviewing large documents. Using multiple LLMs can mitigate, but aren't perfect. I'm generally bullish on LLMs but I would be skeptical fulling relying on it, especially for things I don't fully understand already. If speed is what you're after though, I definitely think they are helpful.
Join the discussion