How to Use AI to Build a TAM/SAM Analysis Without a Research Firm
How to Use AI to Build a TAM/SAM Analysis Without a Research Firm
9 minute readAt some point — in an investor meeting, a strategic planning session, or your own go/no-go decision — someone is going to ask you how big the market is. And the answer needs to be defensible, not directional.
Market sizing has traditionally required access to expensive syndicated research reports or a research firm. AI has meaningfully changed that. Not by generating the numbers for you (that's where founders get into trouble) but by dramatically accelerating the data gathering and synthesis that makes the analysis possible.
Here's what we'll cover:
What TAM, SAM, and SOM actually mean and how to calculate each
The data sources that make the numbers credible
Where AI helps and where it will mislead you
A step-by-step workflow for building a defensible analysis
How to stress-test your numbers before you present them
TAM, SAM, SOM: What Each One Actually Means
These three numbers get used interchangeably in pitch decks. They shouldn't be.
TAM — Total Addressable Market
TAM is the total revenue opportunity if you captured 100% of the market. It sets the outer boundary of how big the opportunity could be. In most pitch contexts, this is the number that shows investors the category is worth paying attention to.
Important: TAM is not your market. It's the universe. Confusing TAM with your actual opportunity is one of the most common credibility errors in early-stage pitches.
SAM — Serviceable Addressable Market
SAM is the portion of the TAM you can realistically reach with your current business model, geography, and go-to-market approach. If your TAM is the global HR software market and you're building for US mid-market companies with 100-500 employees, your SAM is that specific segment.
SAM is where ICP precision matters most. A vague customer definition produces a SAM number that can't be defended under scrutiny.
SOM — Serviceable Obtainable Market
SOM is what you can realistically capture within your planning horizon given your resources, team, competition, and growth rate. This is the number sophisticated investors care most about — it reflects your actual model, not the theoretical ceiling.
SOM should be built up from your unit economics, not down from TAM. If you can close 5 enterprise deals per month, that's your SOM calculation, not a percentage of TAM.
The Data Sources That Make the Numbers Credible
AI cannot generate credible market size numbers from scratch. It can synthesize and apply data from credible sources. Here's where those sources come from:
Government and public databases
The US Census Bureau, Bureau of Labor Statistics, and NAICS industry data are free and credible. For most B2B categories, the Census Bureau's Economic Census provides revenue and establishment counts by NAICS code. This is your TAM foundation.
Industry associations
Most industries have an association that publishes annual market size reports. These are often free or low-cost and carry significant credibility. Google "[industry] association market size report" or "[industry] trade association research."
Publicly available company reports
Public company 10-Ks, investor presentations, and earnings call transcripts often contain market size statements with citations. If a public company in your category says the market is $X billion, that's a citable source.
Free research summaries
Gartner, Forrester, IDC, and IBISWorld often publish free summaries of their paid reports. The full reports cost thousands of dollars, but the executive summaries frequently contain the headline market size numbers you need.
Academic and nonprofit research
Google Scholar, SSRN, and sector-specific nonprofit research organizations often publish market analyses at no cost.
| TAM/SAM/SOM isn't a slide exercise. It's an argument. Every number needs a source it can point to. |
Where AI Helps — and Where It Will Mislead You
Here's the critical distinction founders need to understand:
AI is useful for: identifying which data sources to check for your category, synthesizing multiple market size estimates into a triangulated number, applying segmentation logic to calculate SAM from TAM data, and stress-testing your methodology by generating counter-arguments.
AI will mislead you if you: ask it to generate a market size number directly, ask it what a specific market is worth without providing source data, or treat its output as a citable source. AI-generated market size numbers are either confabulated or drawn from training data that may be years out of date. Neither is acceptable in an investor context.
The workflow that works: you find the primary data sources, you feed them to AI, and AI helps you synthesize and apply them.
Step-by-Step: Building the Analysis
Define your ICP precisely. Your SAM calculation is only as good as your customer definition. Write a specific one-paragraph description of your target customer — industry, company size, geography, job role, and situation.
Find your NAICS code. For B2B markets, the Census Bureau uses NAICS industry classification codes. Find the code(s) that most closely match your category and use the Economic Census or County Business Patterns data to get revenue and establishment counts.
Search for industry reports. Google "[category] market size 2025" and "[industry association] annual report." Look for numbers from credible organizations — not blog posts. Collect 3-5 sources with different estimates.
Use AI to triangulate. Feed your sources to AI with this prompt: "Here are [N] market size estimates for [category] from different sources, ranging from $X to $Y. Analyze the methodology differences that might explain the range and suggest a defensible triangulated estimate with reasoning."
Apply segmentation to calculate SAM. Take your TAM and apply your ICP filters. "If the total US HR software market is $X billion and my ICP is US companies with 100-500 employees (which represent approximately Y% of US businesses per Census data), my SAM is approximately $Z."
Build SOM from unit economics. Don't derive SOM from a percentage of SAM. Build it up: how many customers can you close per month at your current capacity? What's your average contract value? Project 12, 24, and 36 months.
Stress-test with AI. Prompt: "Here is my TAM/SAM/SOM analysis with sources. What are the weakest assumptions in this methodology? What questions is a skeptical investor most likely to ask, and what would the best answer be?"
| Metric | Calculation Method | Key Data Sources |
| TAM | Total market revenue, category-wide | Census, industry associations, public company filings |
| SAM | TAM × ICP segment percentage | Census segmentation data, LinkedIn audience data, industry surveys |
| SOM | Unit economics × growth projection | Your own sales capacity and deal velocity |
Prompts That Work
Data source identification
"I'm building a B2B SaaS product for [describe ICP precisely]. What government databases, industry associations, and publicly available research sources would have the most credible market size data for this category? List each source and what specific data it would contain."
Triangulation
"Here are [N] market size estimates for [category]: [list estimates and sources]. Analyze the differences in methodology and scope that might explain the range. Suggest a defensible triangulated estimate with the reasoning I should use when presenting it."
SAM calculation
"The total US [category] market is approximately $X billion per [source]. My ICP is [describe precisely]. Based on [Census/LinkedIn/industry data], this segment represents approximately Y% of the total market. Help me construct the SAM calculation and identify the weakest assumption."
Stress-testing
"Here is my TAM/SAM/SOM analysis: [describe]. What are the 3 most likely objections an investor or board member would raise about this methodology, and what would be the most credible response to each?"
Frequently Asked Questions
What if I can't find reliable market size data for my category?
For emerging or niche categories, use a build-up approach: estimate the number of potential customers (from LinkedIn audience data, Census establishment counts, or industry membership lists), multiply by average contract value for comparable products, and triangulate against any adjacent market data you can find. A transparent methodology is more credible than a number with a weak single source.
How should I present uncertainty in my market sizing?
Present a range with clear methodology rather than a single number. 'Our TAM estimate ranges from $X to $Y depending on how broadly you define the category. We've used the conservative end of that range for our planning.' This is more credible than false precision.
How often should I update my market sizing?
For early-stage companies, annually or when there are significant category changes. For pitch purposes, make sure your data is no more than 18-24 months old — investors will ask when the numbers are from.
Can I cite AI output as a source in a pitch deck?
No. AI output is not a citable source for market size claims. Every number in your market sizing needs to trace back to a primary or credible secondary source: a government database, an industry report, a public company filing, or peer-reviewed research.
Key Takeaways
TAM, SAM, and SOM are different calculations serving different purposes. Don't confuse them.
AI can't generate credible market size numbers — it can help you find sources and synthesize them.
Credible data comes from government databases, industry associations, and public company filings.
SAM precision depends on ICP precision. A vague customer definition produces a SAM that won't hold up.
Build SOM from unit economics, not from a percentage of TAM.
Need a rigorous TAM/SAM analysis you can defend in front of investors? Praxia Insights conducts market sizing research for founders. |
