Market Research & Client Discovery with AI

Research is where consultants spend the most time and where AI provides the fastest, most measurable wins. A market scan that used to take an associate three days can be drafted in a focused two-hour session. Discovery interview synthesis that used to take a weekend can be done in 30 minutes. This lesson shows you how.

What You'll Learn

• A repeatable AI-powered market research workflow
• How to combine Perplexity, ChatGPT, and Claude for grounded research
• Discovery interview synthesis: from raw transcript to insight
• How to build a competitor scan in under an hour

The AI Research Stack

No single tool is best for research. The strongest consultants combine three:

• Perplexity Pro — for cited primary research. Returns sources, can search the web in real time, includes a "deep research" mode that runs multi-step queries.
• ChatGPT or Claude — for synthesis, structuring, and stress-testing.
• Notebook tool (Claude Projects, ChatGPT Projects, or NotebookLM) — to keep all your gathered evidence in one place across days.

The pattern: gather with Perplexity, analyze with Claude or ChatGPT, persist in a project workspace.

Workflow 1: Market Sizing in 90 Minutes

Suppose your client wants to enter the European HR-tech mid-market. The traditional approach is two analysts and four days. Here is the AI version.

Step 1 — Frame the market with ChatGPT (15 min)

Act as a strategy consultant. Help me define the relevant market for "European HR-tech for mid-market companies (250–2,000 employees)." Give me: (1) a working definition, (2) the 3–5 most credible market segmentation cuts, (3) a list of the data points I would need to triangulate market size. Output as a markdown briefing.

Step 2 — Gather data with Perplexity (45 min)

For each data point ChatGPT identified, run a Perplexity query like:

What is the total addressable market for HR-tech software targeting European companies with 250–2,000 employees, broken down by sub-segment (core HR, payroll, talent, learning)? Cite sources from the last 24 months.

Save each answer with its source links. Use Perplexity's "Deep Research" mode for the harder questions; it will spend 5–10 minutes running multiple sub-queries.

Step 3 — Triangulate with Claude (20 min)

Open a new Claude conversation, paste in the Perplexity outputs, and prompt:

You have the research outputs below from multiple sources, some of which contradict each other. Build a triangulated market size estimate for European HR-tech mid-market. Show: (1) bottom-up estimate, (2) top-down estimate, (3) the assumptions where the two diverge, (4) the most credible single number with a confidence range. Be explicit about uncertainty.

Step 4 — Pressure-test (10 min)

Now act as a skeptical investment-committee partner. Tell me the three weakest assumptions in this market sizing and what additional data we would need to harden them.

You finish with a sized market, a list of caveats, and a clear next-research list — in 90 minutes instead of four days.

Workflow 2: Competitor Scan

For each competitor you want to profile, run this Perplexity prompt:

Profile [Competitor Name] for a strategy assessment. Include: founded year, current ownership, latest revenue (with source and year), employee count, geographic footprint, ICP, pricing model, top 3 product capabilities, top 3 customers if disclosed, recent news in the last 12 months. Use only credible sources and cite each fact.

Then ask ChatGPT to convert the raw outputs into a side-by-side competitor matrix:

I have profiles of 6 competitors below. Build a 6-column comparison table with these rows: revenue, ownership, ICP, pricing model, geographic strength, key product differentiator, biggest weakness. Add a final row: "What this means for our client."

You now have the backbone of a 5-slide competitive landscape section in one sitting.

Workflow 3: Discovery Interview Synthesis

Most consultants leave huge value on the table by treating each interview as a standalone document. AI shines at horizontal synthesis across multiple interviews.

Step 1 — Transcribe

Use Otter, Fireflies, Granola, or your firm's approved transcription tool. Modern transcription is 95%+ accurate.

Step 2 — Clean each transcript

Paste each transcript into Claude:

Clean this raw transcript: (1) remove filler words, (2) attribute speakers consistently, (3) merge fragmented thoughts, (4) keep the meaning unchanged. Do not summarize — I want a full clean transcript.

Step 3 — Code each interview

Read this interview with [role at client]. Extract: (1) verbatim quotes that capture the interviewee's main points (with timestamps if present), (2) themes mentioned, (3) tensions or contradictions, (4) any explicit "asks" of us. Output as a structured markdown document.

Step 4 — Synthesize across all interviews

This is where AI is transformative. Open Claude or ChatGPT, paste all coded interviews (Claude handles long context particularly well), and prompt:

You have 12 coded interviews from a client engagement (a regional retailer's go-to-market diagnosis). Across all interviews, identify: (1) the 5 most consistent themes and which interviewees raised them, (2) the 3 most significant disagreements across roles, (3) the most surprising or counter-narrative quote from each interview, (4) issues raised by frontline staff but missing from leadership interviews. Output as a markdown report.

A senior consultant traditionally does this synthesis over a weekend. With AI it takes a focused afternoon, and the output is often more rigorous because the AI does not get tired or anchor on the loudest interviewee.

Common Pitfalls

• Never put a number into a deck if you cannot click through to the source. Perplexity's citations make this easy; ChatGPT's claims do not.
• Cross-check politically sensitive facts in two tools. If both Claude and Perplexity agree, your confidence rises. If they disagree, dig.
• Do not summarize an interview into bullets and discard the transcript. Keep the cleaned transcripts in your project workspace — you will return to them when writing the report.

Key Takeaways

• Use a stack: Perplexity to gather with citations, Claude or ChatGPT to synthesize, a notebook tool to persist evidence across days.
• Market sizing: frame → gather → triangulate → pressure-test, in 90 minutes instead of four days.
• For competitor scans, profile each competitor with a structured Perplexity prompt, then convert to a comparison table.
• Discovery interview synthesis benefits the most from AI: clean, code, then synthesize horizontally across all interviews.
• Always trace claims back to sources before they enter a client deck.