AI Market Sizing for Founders: TAM That Isn't Fantasy

The slide nobody believes

Every pitch deck has a market sizing slide.

Most of them say something like this:

"The global [category] market is $42 billion. If we capture just 1% of that..."

Every investor in the room mentally switches off.

Not because big markets aren't exciting. Because this number was made up, and they know it.

The standard TAM/SAM/SOM approach works like this: find an industry report, take the headline figure, apply a percentage you feel good about, and present the result as your market opportunity. It is fast, it looks credible, and it tells you almost nothing.

The real question is not how big the total market is. It is: how many specific people have this specific problem, and what will they actually pay today?

That is a very different calculation. And AI tools can now help you do it properly.

What's wrong with top-down sizing

The top-down method is not inherently wrong. The problem is that founders treat it as a starting point rather than a sanity check.

Here is the failure mode. A founder building invoicing software for freelancers finds a report saying the global accounting software market is $11 billion. They declare TAM = $11 billion, SAM = $1 billion (SMBs), SOM = $100 million (they will capture 10%). None of those numbers are grounded in anything.

"We'll capture 10% of the market in year one" is one of the most reliable red flags in a pitch. Real initial adoption rates run at 0.5 to 2%. The founder has not thought about how many freelancers exist in their target geography, what they currently spend on invoicing tools, or whether they are actively looking to switch.

Top-down figures are useful context. They are not a business case.

Bottom-up: starting from the customer

A bottom-up model builds the market estimate from real, countable data.

The formula is simple: number of potential customers × average contract value = addressable revenue.

The hard part is arriving at a defensible customer count. This is where most founders give up and reach for an analyst report. With AI tools, the research that used to take a week now takes a few hours.

The data sources that matter

Here are the places where real market signals live.

Job postings. If 4,000 companies are actively hiring for a role that involves your problem, that is a signal about both market scale and urgency. Search LinkedIn or Indeed for job titles that indicate the pain point. Volume and seniority of postings tell you a great deal about where the market is growing.

LinkedIn company filters. LinkedIn's Sales Navigator lets you filter companies by size, industry, geography, and technology stack. Count the accounts that match your ideal customer profile. That number is far more honest than a percentage of a global figure.

App store data. If competitors already exist, their install counts, review volumes, and rating velocity give you a proxy for category demand. A tool with 50,000 reviews and consistent updates represents a validated segment. A tool with 200 reviews across three years suggests either an early market or a small one.

Search volume. Use Google Keyword Planner or Ahrefs to look at monthly searches for the problem terms your customers would use. Not product names, but problems: "how to manage freelance invoices", "automate expense reports small business". High-volume searches on the problem indicate an active, searching market.

Community size. Subreddit sizes, Discord server memberships, and active forum threads give directional data on how many people are engaged with a problem. Not a precise figure, but useful alongside the others.

How AI compresses the research

AI tools do not replace this research. They compress the time it takes to gather and structure it.

A useful workflow:

1. Ask Claude or Perplexity to summarise industry reports on your category, filtered by geography and company size. Treat the output as a starting point, not a conclusion.
2. Feed your ICP criteria into a prompt and ask it to estimate the addressable account count from public data sources, then cross-check against your own LinkedIn filter count.
3. Load the data into a spreadsheet and use AI to model different pricing and penetration scenarios, so you can see quickly how sensitive your numbers are to assumptions.

Sierra Ventures describes this as using AI as a "TAM co-pilot": you drive the research questions and validate the outputs, the AI handles the aggregation.

What you do not do is ask ChatGPT "what is the market size for invoicing software" and copy the answer into your deck. That will give you a number that sounds authoritative and is grounded in nothing.

A worked example

A founder is building a project management tool for independent architects, specifically addressing the gap between design software and client billing.

Step 1: Count the ICP. LinkedIn filters show approximately 38,000 independent architecture practices in the UK and US combined. Cross-checked against industry association membership data, the figure is plausible.

Step 2: Estimate willingness to pay. Existing tools in adjacent categories charge £25 to £55 per month for similar professional users. Interviews with ten architects confirm they would consider switching for a purpose-built solution at around £40 per month. That price sits comfortably below what they currently spend across fragmented tools.

Step 3: Calculate addressable revenue. 38,000 accounts × £40/month × 12 months = £18.2m ARR if the full ICP converted. That is the TAM.

Step 4: Apply a realistic penetration rate. At 1% penetration over three years, that is £182k ARR. At 5%, roughly £910k. These are buildable targets that can be broken into monthly acquisition goals and used to model a funding ask.

Step 5: Sanity check against top-down data. An industry report puts the architectural software market at £1.8 billion globally. The £18m niche represents under 1% of that, which is credible and defensible. The numbers are in alignment.

The investor in the room can now ask: "How are you going to reach those 38,000 firms?" That is a far more useful conversation than defending a percentage of a global number.

What good market sizing signals

Investors are not looking for a big number. They are looking for evidence that you understand your customer.

A credible market sizing section shows:

• A specific ICP with a countable population
• A price point grounded in comparable tools or direct customer interviews
• An honest penetration scenario with reasoning, not assumptions
• A note on how the addressable market could expand over time

That is a story an investor can stress-test. It shows you did the work.

---

Once you understand the size and shape of your market from the bottom up, you have the foundation for the decisions that follow: which segment to target first, what to charge, and how fast you can realistically grow. If you are still validating whether demand exists at all, the signal hierarchy in our demand testing post covers how to read the evidence before you build. Size your market from the customer up. Use AI to gather the data, not to invent it.