Considering buying a small business?
With Verity
Thinking of just using a chatbot — or reading it yourself? Fair questions ↓
And one limit, stated plainly: Verity can only be as complete as the package — which is exactly why it tells you what the package is missing.
"So do I still need a Quality of Earnings?"
Verity is the analytical half of a QoE, done before you decide whether the deal deserves the other half. And the biggest saving isn't a cheaper QoE — it's the one you don't buy on the deal that died at reconciliation.
One journey from listing to call — read against what you said matters, with every figure traced to its page.
During onboarding you tell Verity exactly what you're looking for — business type, size, revenue, deal structure, risk tolerance. These aren't generic filters: they become your personal acquisition criteria, and every deal is read against them.
PDF, financials, or paste the CIM text directly into Verity.
Every financial line. Every add-back claim. Every assumption the seller made in their favor. Contradictions and red flags surface automatically, scored against your buy box.
A Pass means the deal fits what you told us matters; a Fail means it doesn't. You negotiate with conviction — or walk away early and save the months you'd have spent finding out the hard way.
Verity provides calculation tools and AI analysis — not financial, investment, or legal advice. Better than the alternative. Not the most complete and verified information.
How Verity Protects Your Deal Documents →It's a fair question — ask it before you pay for anything. You could drop all ten PDFs into a chatbot tonight. Or wire an AI up to every document in your deal, hand it an expert prompt, and let it rip. Plenty of buyers do, and the tools that help you do it are multiplying. The AI is genuinely capable: it reads financials, answers in plain English, and never gets tired. So here's the honest version of what happens — and the one thing none of it fixes.
Every one of those tools does the same thing at its core: feed a general-purpose AI more of your deal so its answers get more relevant. That's real, and it's useful. But relevance isn't reliability. The mistake that costs you money isn't the AI knowing too little — it's the AI being fluently, confidently wrong about what it does know. More context doesn't remove that risk. It can hide it, by making the wrong answer read even better.
A chatbot is a brilliant generalist, tuned above all to be helpful and fluent. Diligence is the one job where helpful and fluent works against you. Ask it to make sense of a messy deal and it does exactly what it's built to do: hand you one clean, confident, readable answer. To get there, it will quietly fill the gaps, smooth over the numbers that don't agree, and carry the seller's figures into its summary as if they'd been checked. It isn't lying. It's being helpful. In diligence, that's the risk.
A test workspace held four deals with four asking prices — $2.40M · $1.99M · $540K · $410K — all of them in front of the model. We asked one question with one right answer: which deal has the highest asking price?
Here's the part that matters. It didn't get this wrong because it lacked context — it had all of it, four numbers on one screen. We tried to instruct our way out and got the same wrong answer with better posture. Polish and correctness are independent. The fix wasn't a smarter prompt; it was taking the arithmetic away from the AI entirely. Now the software computes the answer before any AI touches the conversation — the model's job is language, not counting.
Every buyer wiring an AI to their deal documents is running this exact experiment — on a real deal, with real money, and no second run to catch the miss.
Reading documents is the Clerk's job — and every AI does it, yours and ours alike, on the same class of model. Reading was never the hard part. The part that catches a lie is checking every number in every document against every other one, the same way every time, and refusing to pick a favorite when two of them disagree. That's the Guard — and the Guard is software, not AI. It can't be charmed, it doesn't get tired, and it doesn't invent. When your tax return and your CIM state a different SDE, it keeps both and shows you which file each came from, instead of quietly choosing one. Only once the count is settled does the Judge — AI again — explain what it means for your deal. The Judge can flag, contextualize, and advise on what to ask next. It can never add to, remove from, or reword the count.
That's the whole difference. Not a smarter model — a model with a Guard standing over it.
See how the Clerk, the Guard, and the Judge work together →
And yes — Verity is also a real workspace: your deals, documents, and findings stay put across sessions and devices, and every analysis becomes a sourced brief you can hand your lender. Table stakes. We just wouldn't skip them.
The honest answer: for a quick gut-check on a single listing, your own AI is genuinely useful — same model, fast, nearly free. Use it. Verity is for the moment the decision gets real — when several documents, real money on the line, a call you can't take back — when the one thing you can't afford is a confident summary that quietly trusted the seller.
That's the moment you want a Guard.
Analyze your first deal free — then $20 a deal. No subscription.Verity runs a deal through three parts that keep each other honest. Two of them are the AI. The one in the middle — the one that produces the number you act on — is not. That's on purpose.
The AI transcribes exactly what each document says and keeps the document's own words as proof. Blanks stay blank — never filled with a guess about what “a business like this usually” looks like. A quiet listing is made to look neither better nor worse than one that spelled everything out.
Your open items and contradictions come from a fixed set of checks written in software — run identically on every deal, every time. The Guard can't be persuaded and doesn't decide what “seems important”; it checks what it always checks. When two documents state equipment value differently, it keeps both figures and shows them side by side instead of quietly picking one. The same documents produce the same findings, every time — measured in production, not promised on a slide. And if the count ever can't be verified, the Guard says so rather than showing you a confident wrong list.
With the count already settled, the AI reads the same record and adds what models are actually good for: context, patterns, what a figure means for your deal. It can flag something worth a closer look — but it can never add to, remove from, or reword the count.
The Clerk and the Judge are the AI. They read and they comment. But the count comes from the Guard — and the Guard is software code, not judgment. Reading your documents is free; the analysis — the count and the explanation — is what you pay for.
Verity does a lot of work on every deal, but we only charge for the part that runs on artificial intelligence — reading your documents and answering your questions. Our scoring, fit logic, and evaluation tools are ours, and you use them freely.
A few years ago I was evaluating a small business acquisition. A twenty-minute conversation with a broker who'd seen hundreds of these deals reframed everything the seller package had me believing. I needed a co-analyst that would sort every figure into supported, claimed, or missing — and never invent anything to fill a gap. It didn't exist, so I built it.
Truth should not be a luxury in business.
Documents you upload, answers you provide, and questions you ask are processed exclusively within Verity's contracted infrastructure — Supabase for secure storage, Anthropic for AI processing — to generate your results. Your information is never used to train AI models, never shared with third parties, never sold, and never used to benefit anyone other than you.
What goes into Verity stays in Verity's ecosystem — and leaves only when you choose to export it.