Catherine Bell built a custom GPT to research tax strategies for high-net-worth investors, pulling from IRS publications and Treasury guidance with APA citations. The AI can't give exact dollar figures because its data differs from the investor's API, so it only offers qualitative guidance. Human verification of every claim is essential.
This study was created out of the conversation that we had with AI yesterday while I was driving down I-4 from Tampa. Basically, you can see here my recorded message for 4 seconds. Looking, uh, I've got an investor with a multi-million dollar tax liability, let's reverse the agent study is. And so over that 50-minute call with my Deal Hunter by Land Shark, it— that's the— what it basically refused to do was give me actual numbers. Um, I could force it to, but you guys said you guys get your numbers from a different, uh, a different API source. I'm pulling mine from From Lowe's, Home Depot here. See what my sources were. I got the IRS, the Treasury. Oh, at least it's pulling valid, you know, from valid sources. How to claim 100% bonus depreciation. This is an article written, it looks like, up on an exterior website. It looks like a CPA firm, and it's got a pretty decent article about it. So what I did when I created this GPT was I said, because I have a degree in science, I said, you're going to study, you're go— you're going to go out and get research, but you're going to supply me with, with, with the, the references and sources, just like we're, we're supplying an APA format. I wanted it to produce APA-formatted research, right? So I Taught it how to do that. So, and so that's what it does. It goes out and grabs viable sources, brings it back, and then turns around and and tags that into it. So, this is important also to the fact that I'm able to then verify, validate, and verify everything that it's told. Me so that when I do ask the questions, uh, how I'm gonna— answers based on questions I ask, right?
