**Daily Fathom Call Analysis — detect product issues, check GitHub/Linear for duplicates, and surface findings in Slack. Also answer @mention questions about the automation.** --- Trigger 1: Daily Cron (9 AM PT) --- 1. **Fetch recent Fathom recordings** — Use to get all meetings from the last 24 hours. Track which recordings have already been processed using the platform state store (key: `processed_recordings`) to avoid duplicates. If no new recordings are found, post a short "No new calls to analyze" message to **#customer-conversations-pmf** using and stop. 2. **Get full transcripts** — For each new recording, use to retrieve the full call transcript. If a transcript is empty or unavailable, skip that recording and note it in the summary. 3. **Analyze transcripts for product issues** — Use AI to analyze each transcript and extract product issues, frictions, complaints, or feature requests mentioned by customers. For each issue found, extract a **short title**, a **detailed description**, and **relevant quotes** from the transcript. 4. **Check for existing issues in GitHub and Linear** — For each detected product issue, use to search GitHub Issues and to search Linear issues for similar existing issues. Consider an issue "already tracked" if there's a close semantic match. If the search fails for either system, flag the issue as "unable to check for duplicates" and still surface it. 5. **Post a summary thread in Slack** — Use to post a parent message in **#customer-conversations-pmf** (channel ID: **C06J6G6JMCY**) summarizing the day's analysis (e.g., "📞 Analyzed X calls — found Y product issues"). Then for each recording, reply in the thread with what was detected. 6. **Surface issues in threaded replies** — For each detected product issue, post a threaded reply using with the issue details: - If **already tracked**: include a short note like "🔗 Previously detected" with a hyperlink to the matching GitHub issue URL or Linear issue URL (or both if found in both). - If **new/untracked**: include two clickable links — one to create a new Linear issue and one to create a new GitHub issue (pre-filled with the title and description). 7. **Update state store** — Save the list of processed recording IDs to the platform state store so they are not re-processed tomorrow. --- Trigger 2: Slackbot @Mention --- 1. **Receive the @mention message** — When someone @mentions the Decisional slackbot in a channel message, read the incoming message text and channel from the trigger payload. 2. **Answer questions about the automation** — Use AI to answer the user's question about this automation's recent activity. Use the platform state store to look up what was processed (`processed_recordings`), what issues were found (`detected_issues`), and what was posted (`posted_summaries`). Reply in the same channel thread using . If the question is not about this automation, reply politely that you can only answer questions about the Fathom call analysis automation. If the message is unclear, reply asking the user to clarify what they'd like to know.

Call Recording Analyzer Agent

Call Recording Analyzer Agent

Ready to Automate?