Weekly user interview insight repository in Airtable from recordings

Build me a weekly agent workflow that turns raw user research interview recordings into a structured, searchable insight repository in Airtable. The end product should be a thematic, quote-level database that product managers can query, not a flat dump of transcripts.

Trigger: cron, every Friday at 4pm in my local timezone.

Step 1. Use Google Drive's List Files operation to find audio files added in the last 7 days to a designated 'User Interviews' folder. Filter by parent folder ID, by audio mime types (mp3, wav, m4a, flac, webm), and by createdTime within the past week. If no new files are found, end the run quietly.

Step 2. For each new recording, use Google Drive's Download File Content operation to pull the raw bytes.

Step 3. Pass the bytes to Deepgram's Transcribe Pre-recorded Audio operation. Enable diarization and smart formatting so interviewer turns and participant turns are clearly separated, and so numbers, dates, and punctuation render cleanly. Use a high-accuracy model suitable for English interview audio.

Step 4. For each transcript, extract the most quotable participant statements. Skip interviewer prompts, small talk, and filler. Aim for verbatim quotes that capture a clear thought, opinion, or behavior. Trim to roughly one or two sentences each, but do not paraphrase.

Step 5. Classify each quote by one primary theme from this list: pain point, goal, feature request, workaround, positive sentiment. Also tag a coarse sentiment of positive, neutral, or negative.

Step 6. Write the quotes to Airtable using the Create Records operation against a 'Research Insights' table. One row per quote, batching up to 10 records per request. Fields to populate: Quote (long text), Theme (single select from the list above), Sentiment (single select), Participant (text, inferred from the file name or speaker label), Source File (the Drive file name plus a link to the file), Interview Date (date, from the file's createdTime).

Step 7. Also write a short top-level summary per interview to a separate 'Interview Summaries' table using Create Records. Fields: Participant, Interview Date, Source File, Summary (3 to 5 sentences covering who they are, what they care about, and the top takeaways), Top Themes (multi-select of the themes that came up most). One row per interview.

Before creating Airtable rows, ask me for the base ID, the Research Insights table name or ID, and the Interview Summaries table name or ID. Also ask me for the Google Drive folder ID for the User Interviews folder. Validate that the Airtable tables have the expected fields and warn me if any are missing instead of silently failing.

Keep the workflow resilient: if a single file fails to transcribe or classify, log the error and continue with the rest. At the end of the run, log a one-line summary of how many interviews were processed, how many quotes were filed, and any files that were skipped.

Additional information

What does this prompt do?

Finds new user interview audio files added this week to a designated Google Drive folder so you never have to chase recordings manually.
Transcribes each interview with speaker separation so interviewer prompts and participant answers are clearly distinguished.
Pulls the most quotable participant statements, tags each one with a theme like pain point, goal, feature request, workaround, or positive sentiment, and writes them as individual rows in your Airtable research repository.
Adds a short top-level summary per session to a separate table so product managers can scan the week's interviews at a glance.

What do I need to use this?

A Google Drive account with a folder where your team drops user interview recordings.
A Deepgram account for transcription.
An Airtable base with a Research Insights table and an Interview Summaries table ready to receive new rows.

How can I customize it?

Change the schedule. Weekly on Friday afternoon works for most teams, but you can run it daily, biweekly, or right after each interview.
Edit the list of themes. Swap in your own taxonomy, like onboarding friction, willingness to pay, or competitor mentions.
Pick which fields land in Airtable. Add a confidence score, a research question link, or a customer segment column to match your repository.

Frequently asked questions

Do I need a specific Airtable plan for this to work?

Any Airtable plan works as long as you can create records in the base. The prompt assumes you already have a Research Insights table and an Interview Summaries table set up.

What audio formats are supported?

Common recording formats like MP3, WAV, M4A, and FLAC work out of the box. If your team uses a tool that produces a different format, mention it in the prompt and Geni will adapt.

How accurate is the speaker separation?

Speaker separation is reliable for clean two-person recordings. For noisy environments or group sessions, expect some cleanup. You can ask Geni to be conservative and only file quotes when speaker attribution is confident.

Will it duplicate quotes if I re-run it?

The workflow only looks at files added during the current week, so the same recording is not transcribed twice on the normal schedule. If you backfill, ask Geni to check for an existing Source File match before creating rows.

Can I send the summaries to Slack or email instead of Airtable?

Yes. Tell Geni where you want the weekly digest to go and it can post the summaries to Slack, email them, or keep them in Airtable while sending a recap elsewhere.