AI FIELD MANUALCONSULTANTS, FOUNDERS
A source-first market-sizing method - AI structures the framework, but every figure on the slide traces to a named, checkable source, never a model's confident-sounding estimate.
Last reviewed 1 September 2026
An independent consultant is building a market-sizing slide (Total Addressable Market, Serviceable Addressable Market, Serviceable Obtainable Market) for a client's investor deck, due Thursday. Asking an LLM directly - 'what is the TAM for cold chain logistics software in the GCC' - produces a confident, specific-looking number. It is very likely wrong, or at best an artifact of whatever aggregated figures appeared most often in the model's training data, with no way to trace where it actually came from.
A market-sizing number that can't be traced to a source is worse than useless in a client deliverable - it's a liability the moment an investor's own analyst asks for the methodology, which happens routinely in real due diligence.
The task isn't 'get a number.' It's building a number with a visible, checkable calculation and named sources behind each input, which is a different and slower process than asking a chat model for the answer.
Build the market size bottom-up from named, sourced inputs (industry reports, government trade statistics, company disclosures) rather than top-down from a single vague figure. Use the LLM to structure the TAM/SAM/SOM framework, identify what specific data points the calculation needs, and help draft the narrative around the sourced numbers - never to supply the numbers themselves.
Every number that ends up on the slide should have a name next to it in the working file: which report, which page or table, what year. If a needed input genuinely can't be found from a real source in the time available, the honest move is a clearly labeled estimate with the stated assumption, not a number presented with the same confidence as a sourced one.
A bottom-up calculation (number of addressable customers × average realistic contract value, both sourced separately) is inherently more checkable than a single top-down market-size figure, because each input can be verified independently and the arithmetic connecting them is visible - an investor's analyst can push on any one input rather than having to accept or reject the whole number at once.
LLMs are useful here specifically because structuring a TAM/SAM/SOM framework, identifying what data the calculation actually needs, and drafting narrative around confirmed numbers are all tasks that don't require the model to know a true fact about the world - they require it to organize and write, which is what it's actually reliable at.
Write the exact scope: geography, customer segment, product category. "Cold chain logistics software for UAE-based food distributors with 50+ vehicle fleets" is sizeable; "logistics software in the Middle East" is not - ambiguity here makes every downstream number unverifiable.
Ask Claude or ChatGPT to break the TAM/SAM/SOM calculation into its component inputs for this specific market definition - e.g. number of companies matching the segment, average relevant spend per company, expected realistic market share. This is a structuring task, not a data-lookup one.
For each component identified, find a real source: government trade/industry statistics (UAE Federal Competitiveness and Statistics Centre, Dubai Chamber reports), industry association data, named market-research reports, or public company disclosures. Write down the source and date next to every number as you find it.
Perplexity or a browsing-enabled Claude/ChatGPT can speed up finding candidate sources for a specific input - but every candidate source it surfaces gets opened and confirmed, exactly as in the competitive-brief workflow, before the number is used.
Calculate TAM, SAM, and SOM from the sourced inputs in a visible spreadsheet, not inside a chat conversation - the calculation itself should be an artifact the client (or their analyst) could re-derive from the same sourced inputs.
Once every number is sourced and calculated, ask the model to draft the slide's supporting narrative from the finished, sourced figures - explicitly instructed to state each source inline and add no additional figure not already calculated.
Accelerating the search for candidate sources for a specific market-sizing input, verified before use.
www.perplexity.ai (opens in a new tab)Structuring the TAM/SAM/SOM framework and drafting narrative around already-sourced, calculated figures.
claude.ai (opens in a new tab)Genuinely reliable, current, specific market data for a narrow segment (especially in emerging or fast-moving markets) is often simply not available from any public source in the time you have. The honest deliverable in that case states the gap and shows a clearly labeled, assumption-based estimate - not a number dressed up to look as sourced as the rest of the slide.
A bottom-up, sourced calculation is only as good as the input sources themselves - an industry report with its own questionable methodology produces a sourced-but-still-shaky number. Naming the source is necessary, not sufficient; a quick sanity check on how the source itself derived its figures is worth the extra few minutes for any input carrying real weight in the final number.
This method takes longer than asking an LLM for a number directly - that's the trade a client-facing deliverable should be making deliberately, not something to shortcut when the deadline is tight. If time truly doesn't allow proper sourcing, say so to the client rather than delivering an unsourced number with a sourced one's confidence.
Advertisement
A consultant needs a defensible TAM/SAM/SOM slide for a cold-chain logistics software startup's investor deck.
Market defined precisely: cold chain logistics software for UAE-based food distributors and 3PLs operating 50+ vehicle fleets.
Claude broke the calculation into inputs: number of matching companies in the UAE, average annual logistics-software spend per company, realistic achievable market share in year 3.
Number of matching companies sourced from a Dubai Chamber of Commerce industry report (named, dated, page cited); average spend sourced from a named market-research report's per-seat pricing benchmark, cross-checked against two competitors' public pricing pages.
TAM/SAM/SOM calculated in a spreadsheet from the two sourced inputs plus the founder's own stated go-to-market assumption for year-3 share (clearly labeled as an assumption, not a sourced figure).
Claude drafted the slide narrative from the finished spreadsheet, with each figure's source named inline in the speaker notes for when an investor's analyst asks.
A market-sizing slide where every number traces to a named source or a clearly labeled assumption, built to survive the specific question - 'where did that come from' - that unsourced AI-generated market sizes routinely fail.
Nakoda editorial · last reviewed
This entry describes a workflow Nakoda recommends - it is not a claim about how any named tool behaves in every case, and it is not paid placement. Spotted something out of date? Tell us.
Workflow
How to build a defensible AI-competitive-landscape brief for a board meeting without an LLM quietly inventing a competitor's product, partnership or funding round.
Workflow
A structured, AI-assisted way to turn a stack of raw customer discovery transcripts into honest themes, using a real qualitative-research method rather than asking a chatbot to "summarize the feedback."
Pipeline
A real evaluation methodology - golden dataset, LLM-as-judge with a written rubric, and a regression gate - for teams about to ship an AI feature, instead of shipping on "it looked right in testing."
