Synthesize customer feedback from Slack, Gmail, and Zendesk every week

I want an agent workflow that runs every Friday at 3pm in my timezone and synthesizes the week's customer feedback from Slack, Gmail, and Zendesk into structured records, a theme readout, and Linear backlog drafts.

On each run, gather the inputs:

1. From Slack, fetch the last 7 days of messages in the customer-facing channels I specify at setup (for example #customer-feedback, #support-escalations, #design-partners). Use Get Channel History for each channel, and pull Get Thread Replies for any thread with more than one reply so you capture full context.

2. From Gmail, use List Messages with a date filter for the last 7 days and a label or query I configure at setup (for example label:product-feedback OR from:customers@). Skip newsletters, calendar invites, and automated notifications.

3. From Zendesk, use Search Tickets to fetch tickets updated in the last 7 days that match a view or query I configure (for example, tickets tagged product-feedback or feature-request, or any ticket in a specific support group). For relevant tickets, pull List Ticket Comments so you have the full conversation, not just the subject.

For every distinct piece of feedback you find across these sources, create a structured record by calling Notion's Create a Page in the feedback database I point you at. Each record must include: a one-sentence summary, the verbatim quote (trimmed but not paraphrased), the customer name or workspace if known, the source (Slack, Gmail, or Zendesk), a direct link back to the original message or ticket, a feedback type (bug, feature request, UX friction, churn signal, praise, other), a product area, a confidence score (low, medium, high) for how clearly the customer expressed the need, and any open questions you would want a PM to follow up on.

Before creating a Notion record, use Query a Data Source on the feedback database to check whether a near-duplicate already exists for the same customer in the last 30 days. If it does, update the existing record with the new evidence link instead of creating a new one.

After all records are written, cluster the week's records into recurring themes. A theme should have at least 2 independent customers behind it. For each theme, write: theme name, plain-language description, count of supporting records, list of customers, severity (low, medium, high), and 2 to 4 verbatim quotes as evidence.

For every theme with 3 or more customers and medium or high severity, draft a Linear backlog item using Create Issue in the team I specify at setup. The issue title should read like a real product brief ("Customers cannot bulk-archive projects from the sidebar"), the description should include the problem, evidence (links back to the Notion records), open questions, and a suggested next step. File the issue in a triage workflow state so a PM has to promote it. Before creating, use Search Issues in Linear to check for an existing similar issue; if you find one, post a comment with the new evidence instead of opening a duplicate.

Finally, post a readout to the Slack channel I specify (for example #product-weekly). The readout should include: a one-paragraph summary of the week, the top 3 to 5 themes with their counts and severity, links to the Linear drafts you created or commented on, links to the Notion records, and a short "open questions" section listing things you were unsure about. Use Send a Message in Slack and format it cleanly with sections and bullets.

Important guardrails: never paraphrase a customer quote in a way that changes its meaning; always preserve a source link on every claim; if a signal is ambiguous, mark its confidence as low and call it out in the open questions section rather than guessing; and do not delete or modify existing Notion records or Linear issues you did not create on a previous run of this workflow.