Skip to content
agentusehub
ResearchBeginnerEducation

Turn customer interviews into a themes database

Descript transcribes each discovery call, Claude tags pains, requests and quotes, and Notion keeps a themes database the whole team can search.

Time saved
4h / week
Setup time
45 minutes
Running cost
$32/mo
Cost range
$25–100/mo

The problem it solves

Edtech product teams run discovery calls with teachers and school admins, then the recordings sit in a folder. Insights live in one person's head or a messy doc, so the same questions get asked again. When it is time to prioritise the roadmap, nobody can say how many people actually raised a problem.

How it flows
  1. Upload recording
  2. Transcribe
  3. Tag insights
  4. Merge into Notion

Step-by-step setup

  1. Transcribe in Descript

    Import the recording into a Descript project with speaker detection on. Rename the speakers to Interviewer and Participant, then export the transcript as plain text.

  2. Create two Notion databases

    Make an Interviews database (Participant role, School type, Date, Recording link) and an Insights database (Insight, Type, Quote, Theme, Interview). Connect them with a relation property on Interview.

  3. Set up a Claude Project

    Create a Claude Project called Interview Synthesis. Add your current theme list and a one-page product overview as project knowledge, and paste the extraction prompt into the project instructions.

  4. Extract insights per call

    Paste each transcript into a new chat in the project. Save Claude's CSV output as a file and use Merge with CSV on the Insights database, then set the Interview relation for the new rows.

  5. Review themes monthly

    Group the Insights database by Theme and sort by count. Run the synthesis prompt on a CSV export to decide which NEW themes to keep, then update the theme list in the project knowledge.

Example prompts

Extract insights from a transcript
Below is a transcript of a discovery interview with a {{participant_role}} at a {{school_type}}. Speakers are labelled Interviewer and Participant.

Extract every distinct insight from the Participant only. Return CSV with the columns Insight, Type, Quote, Theme.

- Insight: one sentence in your own words.
- Type: Pain, Request or Praise.
- Quote: the participant's exact words, under 30 words.
- Theme: pick from the theme list in the project knowledge. If none fits, write NEW: followed by a short name.

Ignore small talk, and ignore anything the interviewer suggested that the participant did not confirm.

Transcript:
{{transcript}}
Monthly theme synthesis
Attached is a CSV export of our Insights database for {{month}}. Group the insights by Theme and count them.

For the five largest themes, write two sentences on what participants are struggling with and include the two strongest quotes.

Then list any NEW: themes that appear three or more times, and recommend whether each should become a permanent theme or be merged into an existing one.
Research · Intermediate
  1. Monday 7am
  2. Search 50 sources
  3. Diff vs last week

Weekly competitor brief from 50 sources

A CrewAI crew checks 50 competitor pages, changelogs and news sources every week and sends a short brief covering only what changed.

CrewAI · Perplexity · Claude
Sales · Beginner
  1. Form submit
  2. Fetch company site
  3. Claude scores fit

Auto-qualify inbound leads before a rep sees them

Score every demo request against your ICP, update HubSpot and ping the right rep in Slack within a minute of the form submit.

n8n · Claude · HubSpot · Slack
Customer Support · Intermediate
  1. New ticket webhook
  2. Claude classifies
  3. Draft from help docs

Triage support tickets and draft first replies

Every new Zendesk ticket gets a category, priority and sentiment tag plus a draft reply in an internal note before an agent opens it.

Zendesk · n8n · Claude