Size Your Market and Price Your Product With AI, in an Afternoon

Two numbers get hand-waved in almost every early deck: how big the market actually is, and what the thing should cost. Founders sprint through both because both are genuinely hard – and because getting them wrong in public is embarrassing. So you get the slide with “$47B market” in 90-point font and no arithmetic behind it, and a pricing page that’s really just three tiers you invented on a Tuesday.

The problem is that these are exactly the two numbers a sharp investor or a serious customer will push on first. A market number you can’t defend torches your credibility for the rest of the meeting. A price you picked by vibes leaves money on the table or scares off the buyers you needed most.

Here’s the good news: with the current crop of AI tools you can get a genuinely defensible first pass on both in a single afternoon – not a finished, board-grade artifact, but something real enough to build on. Here’s the playbook, plus an honest account of where generic AI quietly falls apart if you lean on it alone.

First, the trap: why a raw chatbot isn’t enough

The obvious move is to open ChatGPT or Claude and ask for a TAM. GPT-5.5 and Claude Sonnet 5 are both strong, fast, and cheap, and they’ll happily produce a tidy market-sizing paragraph in seconds. That paragraph is a good starting sketch and a terrible final answer, for three specific reasons.

• It invents sources. Ask for the citation behind “$12B and growing 18% annually” and you’ll often get a report title, a firm, and a year that don’t line up – or don’t exist. GPT-5.5 shipped with lower hallucination rates in law, medicine, and finance, which helps, but “lower” is not “zero,” and market data is exactly the domain where a confident wrong number is worse than no number.
• It won’t reconcile. A real TAM is built two ways – top-down (start from the whole industry, cut to your slice) and bottom-up (number of buyers times what each pays) – and the two rarely match on the first try. The reconciliation, the arguing between the two methods until they agree within reason, is the actual work. A single prompt gives you one method, presented with total confidence, and no tension between approaches.
• It doesn’t know how confident to be. Everything comes out in the same assertive voice, whether it’s a hard census figure or a wild guess about willingness-to-pay. There’s no signal for which numbers to trust and which to flag.

None of this means the raw chatbot is useless. It’s a fine way to explore the shape of a market before you commit. It’s just not the thing you put in front of a partner at a fund. So use it for the sketch, then move to tools built for the reconciliation.

The afternoon playbook

Block three or four hours. The sequence matters – market first, because your pricing sanity depends on knowing who and how many you’re actually selling to.

1. Write down your slice in one sentence. Not “SaaS” – “monthly subscription scheduling software for independent physiotherapy clinics in the US and UK.” Every number downstream is only as good as this sentence. Vague input, garbage output.
2. Do a naive first pass in a chatbot. Ask Claude or ChatGPT for a rough top-down and bottom-up sizing of that slice. Read it for shape and vocabulary – what segments exist, what the buyers are called, what the obvious revenue drivers are. Treat every number as a hypothesis, not a fact.
3. Build the real TAM/SAM/SOM properly. This is where you run both methods in parallel, force them to reconcile, attach sources you can actually click, and mark your confidence on each input. Do it in a spreadsheet if you have the patience, or hand it to a tool built for exactly this (below).
4. Pull live competitor pricing. Before you price anything, gather what the market already charges – tiers, headline numbers, what’s gated behind “contact sales.” This is your reality anchor.
5. Layer willingness-to-pay on top. Competitor prices tell you the going rate; they don’t tell you what your buyer will pay for your wedge. Combine the two into an actual pricing architecture – tiers, anchors, what goes in each, where the upgrade pressure lives.
6. Stress-test the whole thing. Feed your SOM and your price into a back-of-envelope check: does the revenue math survive contact with your realistic CAC and payback? If your beautiful TAM implies a SOM you can’t afford to acquire, you learned that in an afternoon instead of a year.

The tools that do the heavy lifting

Steps 3 through 5 are where generic AI stops being enough and purpose-built tooling earns its keep. Two VentureVerse apps map onto this playbook almost exactly.

Market sizing: the reconciliation problem, solved

The Market Sizing Calculator is built around the specific failure mode above. Instead of one method delivered with false confidence, it runs top-down and bottom-up research in parallel and reconciles them into a single defensible TAM/SAM/SOM – the argument-between-methods that you’d otherwise have to referee yourself. It attaches confidence scores, so you can see which inputs are solid and which are educated guesses, and it produces an investor-ready export you can drop into a deck without reformatting.

The honest framing: this is the difference between “a market number” and “a market number you can defend when someone pushes.” It’s the natural founder pick here precisely because it does the two things a raw chatbot won’t – reconcile the methods and grade its own certainty.

Pricing: from competitor scrape to real architecture

The GTM Pricing Decoder handles steps 4 and 5 together. It builds a full pricing architecture from live competitor pricing plus willingness-to-pay data – not “here are three tiers,” but a reasoned structure: where to anchor, what belongs in each tier, where the natural upgrade pressure sits. That’s the part founders most often skip, and the part that quietly determines whether your revenue model works.

Pair it with the Meridian for step 6 – it runs CAC, LTV, payback, and margin against benchmarks, so you can immediately see whether the price you just designed and the market you just sized actually cohere into a business. This is the check that turns two impressive slides into one honest one.

Honest alternatives

You don’t strictly need any of this. If you have the time and the discipline, a raw model plus a spreadsheet gets you most of the way – the models are good enough now that Claude Sonnet 5 or GPT-5.5 will draft the structure and even the formulas, and you supply the judgment and the real citations. The catch is that “you supply the citations” is the whole job, and it’s the part people skip when they’re busy. That’s the trade: DIY is free and flexible but puts the reconciliation and the fact-checking entirely on you.

If your market is genuinely novel – a category that doesn’t exist yet – no tool will conjure data that isn’t out there. In that case the honest move is a bottom-up build from first principles (units, buyers, price) and explicit assumptions you’re willing to defend, rather than a top-down number borrowed from an adjacent market. And if you want to pressure-test the story before you tell it, running your assumptions through a tool like Risk Matrix – which writes the post-mortem of your failure before it happens – is a bracing way to find the market-and-pricing assumptions you’re quietly betting the company on.

The bottom line

Market sizing and pricing feel like research projects because founders treat them as one-shot questions and reach for a chatbot that answers with confidence it hasn’t earned. The reframe is simpler: both are reconciliation problems. A defensible TAM is two methods forced to agree; a real price is competitor reality reconciled with what your buyer will actually pay. Do that reconciliation properly – by hand or with tools built for it – and you can walk out of an afternoon with two numbers you’ll happily defend under questioning, instead of two you’re praying nobody asks about.

Explore the VentureVerse apps (start with the Market Sizing Calculator and the GTM Pricing Decoder) and Get The Brief.