Module 3: AI-Native Market Research — AI Strategy for Business Leaders

What this module covers

Design and run an AI-assisted market-sizing, pricing, and qualitative-research exercise, and know how to validate AI-generated findings rather than just trusting them.

Lessons in this module

3a — Bottom-up market sizing
Building a TAM → SAM → SOM funnel from public directory/membership data.
3b — Pricing psychology in practice
Three real behavioural mechanisms: anchoring, framing, and loss aversion, each illustrated with an actual pricing decision.
3c — Validating AI research: triangulation over trust
Where AI-assisted research needs independent checking — catching a vendor's unverified claim, cross-querying multiple models, and a technical audit that sharpened rather than confirmed a hunch.

Deliverable

Try this: Run a scoped AI-assisted TAM/SAM/SOM exercise on your own market and present the funnel with sources.

Check your understanding

1In the Solstice bottom-up sizing exercise, what was the SAM derived from?

A paid market research report
Public professional-body directory and membership data
A survey of venture capitalists
Government census data alone

2Why was a banking-details request reframed as “connect your account so clients can pay you directly”?

To make the request sound more technical
Because the same action reads very differently depending on whether it's framed as a fee/trust ask or a service enablement — a framing effect
To comply with a legal requirement
It has no effect on user behaviour either way

3What is the main risk this module identifies in AI-assisted qualitative research?

It's too slow to be useful
Synthesising findings into a summary without preserving the raw source data makes later re-analysis impossible
It only works for B2C markets
It cannot process text data at all

4What does cross-querying multiple AI models on the same question help protect against?

Increased AI subscription costs
Treating a single model's confident-sounding answer as settled fact, when agreement or divergence across models is itself useful signal
The need for any human review
Market sizing errors specifically