Build a customer interview research repository in Notion from Granola notes

Build me an agent workflow that turns every customer interview I take in Granola into a structured record in our Notion research database, and posts a short highlights digest in our research Slack channel so the team sees it.

Trigger: poll Granola for new notes (event type: new_note), scoped to a specific Granola folder I will configure (our Customer Interviews folder). I will provide the folder name or ID at setup. Skip any note whose meeting title does not look like a customer interview, using simple heuristics like the presence of an external company name, or a title pattern such as 'Interview', 'User Research', 'Discovery', or 'Customer Call'. When in doubt, skip rather than file a noisy record.

For each interview note that passes the filter, the agent should:

1. Call Granola Get Note for the triggering note ID with include=transcript so we have both the summary_markdown and the full transcript to work with.

2. From the summary and transcript, extract the following structured fields: interviewee name, interviewee role/title, company, interview date, key pain points (as a list), themes (as a list of tags), top 1 to 3 surprising or quotable lines (verbatim, with attribution), and any explicitly requested features or product asks. If a field is genuinely missing from the note, leave it blank rather than hallucinating.

3. Call Notion Create a Page with the parent set to our research database (I will paste the database/data source ID at setup). Map the extracted fields onto the database properties: Interviewee (title), Company, Role, Interview Date (date), Themes (multi-select), Pain Points (multi-select or rich text), Feature Requests (rich text), and a Granola Link property pointing back to the original note URL. Put the full Granola summary_markdown into the page body, followed by a Surprising Quotes section with the verbatim quotes, and an Action Items section if Granola surfaced any. If a property does not exist on the database, skip it gracefully.

4. Call Slack Bot Send a Message to post a short digest to the research channel I designate at setup. The message should include: interviewee name and company, 2 to 4 bullet highlights (top pain points and themes), one standout quote, and a link to the new Notion page. Keep it under ~10 lines so it reads cleanly in channel.

Other behavior I want: be idempotent (do not double-create the Notion page if the same note fires twice; check by Granola note ID stored in a property like 'Granola Note ID' on the database before creating). If Notion create fails, do not post to Slack. If Slack post fails after a successful Notion create, log the error but do not retry the Notion create. Use the Granola summary_markdown verbatim for the body, do not rewrite it.

At setup time, ask me for: the Granola folder to watch, the Notion research database/data source ID, the Slack channel ID for the digest, and any custom title keywords I want to treat as 'this is an interview'.