中文 Book a Workshop
AI workflow · Marketing · Research

Customer Interview Analysis with AI

AI-assisted interview analysis reads every transcript the same way and returns themes, each backed by verbatim quotes and a count of how many interviews mentioned it — so the team argues about evidence, not about who remembers what.

TeamMarketing
LevelIntermediate
ToolsClaude · ChatGPT
How it usually goes
  • Five people each remember a different interview.
  • The loudest quote becomes "what customers want".
  • Transcripts sit unread in a folder.
The workflow · 5 steps
  1. 1

    Anonymise, then load

    Replace names and companies with labels (Customer A, B…). Check your AI tool is approved for customer data before pasting anything.

  2. 2

    Code one transcript at a time

    For each: problems mentioned, workarounds, words they used for the problem, what they tried before. Verbatim quotes only.

  3. 3

    Merge into themes with counts

    Ask for themes across all coded transcripts, with the number of interviews that mention each and two quotes per theme.

  4. 4

    Ask what does not fit

    Request the statements that contradict the main themes. Minority signals are where models smooth things over.

  5. 5

    Spot-check quotes against transcripts

    Open three quotes at random and find them in the original. If one is paraphrased or missing, rerun with stricter instructions.

The prompt · copy and paste
Below are [N] anonymised customer interview transcripts.

1. For each transcript: list problems, current workarounds, and the exact words the customer used. Verbatim quotes only, in quotation marks.
2. Across all transcripts: group into themes. For each theme give the count of interviews that mention it (e.g. 4 of 7) and two verbatim quotes with the customer label.
3. List statements that contradict or do not fit the themes.
4. Do not recommend anything. Analysis only.

If a quote is not word-for-word in the transcript, do not use it.

TRANSCRIPTS:
[paste]
Example output · fictional company
Theme 1 — Reporting takes a full day each week (5 of 7)
"I lose every Friday to the same spreadsheet." — Customer C
"Nobody trusts the number until I have rebuilt it by hand." — Customer F

Does not fit: Customer B says reporting is "fine" and the real problem is onboarding new hires.

The full reusable version, with input fields and customisation notes: Customer Interview Analysis Template.

Before it goes out · a person checks

Review checklist

  • Three random quotes were found word-for-word in the transcripts.
  • Counts match: recount one theme by hand.
  • Contradicting statements are listed, not hidden.
  • No customer can be identified from the output.
Don’t use it for

Where this workflow stops

  • Fewer than four interviews — read them yourself; there is no pattern to find yet.
  • Deciding what to build. The output is evidence for a decision, not the decision.
  • Any tool not approved for customer data.

Part of AI for Marketing. Not sure your team is ready for workflows like this? Take the AI Readiness Assessment — 10 questions, no email. Want your team to build this on its own work? See how training works.

Related

Workflows that pair with this one

All workflows →
Marketing

Competitive Research with AI

Build a side-by-side competitor comparison from their public pages: positioning, pricing, claims and gaps, every cell sourced.

BeginnerChatGPTClaude
Open workflow →
Marketing

Content Repurposing with AI

Turn one long piece — a webinar, a guide, a talk — into a LinkedIn post, an email and a short script without losing the point.

BeginnerChatGPTClaude
Open workflow →
Leadership

Executive Brief Creation with AI

Build a one-page weekly brief for leadership from team updates: what changed, what is at risk, what needs a decision.

IntermediateCopilotChatGPT
Open workflow →

Bring one task.
Leave with it running.

Book a 30-minute call. You’ll have a recommended format and a written quote within 24 hours.

Book a Workshop

Or email [email protected] · text TRAINING to (469) 425-4142