Earlier this year we asked how you’re using AI in your search. Two things crossed my screen this week that answer it well. The first is a Searchfunder post from Emile Petrone, a solo searcher, on the five ways he uses AI. The first four are what you’d expect: daily listing alerts, reusable deal analysis prompts, Claude’s Excel extension to standardize messy seller financials, and document generation. The fifth is the interesting one. Instead of reacting to listings faster, he built a database of small businesses from PPP data, estimated SDE for each, and filtered millions of companies down to a few hundred owners he can contact directly. His post also links to 35 prompts you can customize. Worth checking out here. The second is something we’ve been doing at EC. I posted a prompt on LinkedIn that turns Claude Code into a buy-side analyst for HVAC, plumbing, and other trades deals. Give it a business name, website, and owner, plus the seller’s claimed revenue and SDE, and it pulls from public sources: entity filings and liens, licensing, permits, Google reviews, ad libraries, job postings, PPP data, and more. It then triangulates a revenue estimate and scores the seller’s claims as supported, contradicted, or unverifiable. Two things have made the difference for us. One is giving Claude real tools, like an Apify API key, so it can collect data instead of guessing. The other is the rule that every finding gets tagged Verified, Inferred, or Unverified. AI will happily sound confident about things it hasn’t confirmed. We still do plenty of manual work, but this keeps surfacing things worth asking about. The full prompt is in my LinkedIn post. Partner Perspective: Caleb Basile, QoE Prep: How Far Off Can the Seller’s EBITDA Be Before the Deal Dies? I recently sat down with Jake Demi, a bookkeeping specialist who helps small business owners get their books in order. We talked about what makes books trustworthy and what a QoE provider is afraid to find when he opens them. I wrote up the conversation on my Substack (link here), but one part of it deserves its own spotlight. Every seller thinks their numbers are clean. Almost every broker thinks the adjusted EBITDA in the CIM is right. Jake sees the same thing from the bookkeeping side: owners who swear their books are simple, until someone looks at the balance sheet. Then a quality of earnings report comes back with a different number. What happens next depends on how big the gap is. In my experience, there’s a pattern. If the seller’s proposed adjusted EBITDA is within 10% of mine, the deal usually just goes through. A small difference is normal, and both sides can live with it. If it’s 10% to 35% apart, there’s a negotiation. It might be price, it might be terms, but there is always a conversation. If it’s more than 35% apart, the deal usually dies. At that point the buyer and seller aren’t looking at the same business, and there’s no good way to bridge it. The goal of a QoE isn’t to drop the price. It’s to find the true earnings. I don’t negotiate for my clients, but I do hand them the three things worth raising, and the broker can then explain to the seller why the number has to move. I’ve seen prices drop in half after a QoE, and I’m paid either way, which keeps me honest. If I only got paid when the deal closed, I’d have a reason to look the other way. If you’re a seller, find the gap before the buyer does. If you’re a buyer, a QoE tells you which of these three conversations you’re about to have.redacted