TemplatesResearchWyse

AI Customer Research Brief Template

Customer research stays valuable. AI synthesis makes it 10x faster. This is the brief format that gives AI enough structure to produce usable findings instead of pattern-matching noise.

6 days for a full research cycle (vs 6 weeks baseline)Difficulty: mediumFree, no signup

Why this matters

The template, step by step

  1. 01

    Define the decision the research informs

    What product decision will this research influence? 'Should we build feature X?' 'Why are users churning at month 3?' 'What's the right pricing for the new tier?' If you can't name the decision, you don't need the research yet.

  2. 02

    Pick the participant profile

    Who do you need to interview? Specific titles, company sizes, behavior patterns. Don't talk to everyone; talk to 8-12 people who match the profile that will inform the decision.

  3. 03

    Write 10-15 interview questions

    Open-ended, behavior-focused, not opinion-focused. 'Tell me about the last time you...' beats 'Would you use this?' AI synthesis works better on behavioral answers than opinion answers.

  4. 04

    Conduct and transcribe interviews

    45-60 minutes per interview. Record with permission. Transcribe automatically (Otter, Fathom, Wysera). Don't try to synthesize during the interview; capture and synthesize after.

  5. 05

    Feed transcripts to AI with the brief

    Wyse (or Claude, GPT) takes transcripts plus the brief context (the decision, the profile, the questions). Outputs: top patterns across interviews, contradictions, surprising findings, specific quotes per pattern.

  6. 06

    Translate to decisions

    Final section: based on the findings, what's the recommendation for the decision in step 1? Include uncertainty: 'high confidence on X, medium on Y, low on Z.' Decisions made on low-confidence findings will need follow-up research.

Customer research brief
RESEARCH BRIEF

1. DECISION THIS INFORMS
   ____________

2. PARTICIPANT PROFILE
   Titles: ____________
   Company size: ____________
   Behavior signals: ____________
   Sample size: ____ (target 8-12)

3. INTERVIEW QUESTIONS (10-15)
   Open-ended, behavioral
   Q1: Tell me about the last time you ____
   Q2: Walk me through how you ____
   Q3: ____
   ...

4. SYNTHESIS PROMPT (paste with transcripts)
   "Given these [N] transcripts and the brief above, identify:
   - Top 5 patterns across interviews
   - Contradictions or splits in the data
   - Most surprising findings
   - 3-5 verbatim quotes per pattern
   - Implications for the decision in section 1
   - Confidence level for each implication (high/medium/low)"

5. FINDINGS (AI-synthesized + your verification)
   Pattern 1: ____________
     Evidence: ____________
     Quotes: ____________
   Pattern 2: ____________
   ...

6. RECOMMENDATION (your work)
   For the decision in section 1:
   Recommend: ____________
   Confidence: [high / medium / low]
   Required follow-up if low: ____________

Common mistakes

Skip the manual work

Let Wyse run this template on autopilot.

Wyse drafts every input, every personalization, every follow-up in your brand voice. You approve before anything goes live.

Questions

How is this different from traditional customer research?

Same interview protocols, same participant profiles. The synthesis step is 10x faster with AI. A project that took 6 weeks (1 week setup, 2 weeks interviews, 3 weeks synthesis) becomes 6 days (1 day setup, 2 days interviews, 3 days synthesis plus translation).

Can AI conduct the interviews?

Not yet, reliably. Humans are still better at follow-up probes, reading nonverbal cues, and pivoting when the interview takes an unexpected turn. AI is great at synthesis and at drafting the interview guide.

How many transcripts before AI synthesis is reliable?

Minimum 5, ideal 8-12. Below 5, AI tends to over-generalize from individual answers. Above 15, synthesis quality plateaus while cost rises.

What about privacy and PII in transcripts?

Redact PII before AI processing (names, company names, identifying details). Wysera's research agent runs inside your tenant with PII redaction by default. Customer data never trains public models.

Can this replace user testing?

No. This is for generative research (what should we build, why are users behaving this way). User testing remains the right tool for evaluative research (does this design work, does this flow convert).

Related templates