Sync Fireflies Meeting Notes to HubSpot CRM

Fireflies meeting-completed webhook → GraphQL transcript fetch → Gemini extracts summary & CRM updates → HubSpot writes notes, tasks & properties → optional Slack alerts

Who

Sales or customer-success teams already using Fireflies for meetings and HubSpot for CRM

You get

After each meeting, automatically write transcript summaries, intent scores, and follow-up tasks into HubSpot contact/company/deal records, with Slack alerts for high-intent or error states

You need

n8n · hubspot · gemini · slack

Steps

1. Connect credentials and configure the workflow

   Create and connect four credentials in n8n: Fireflies (Header Auth), HubSpot (App Token), Google Gemini API, and Slack (OAuth2). Open the Workflow Configuration node and set INTERNAL_EMAIL_DOMAINS (e.g. yourcompany.com, comma-separated for multiple), UNKNOWN_CONTACT_ACTION (skip/notify/create), SLACK_CHANNEL, AI_MODEL (default gemini-3.1-flash-lite), and toggles for CREATE_COMPANY_UPDATE, CREATE_DEAL_UPDATE, CREATE_FOLLOWUP_TASK, and SEND_SLACK_ALERT.

2. Configure the Fireflies webhook trigger

   In the Fireflies console, set the 'meeting completed' webhook URL to this workflow's Webhook URL with path fireflies-meeting-completed and method POST. Once saved, Fireflies will automatically push each completed transcript.

3. Fetch meeting details and validate / deduplicate

   After trigger, the Get Meeting Details node pulls the full transcript (sentences and attendees) via the Fireflies GraphQL API. The Validate & Normalize Meeting node checks the payload, deduplicates against staticData.processedMeetings, skips meetings with no transcript, and merges config with meeting data.

4. Fetch HubSpot writable properties and filter external attendees

   The Fetch HubSpot Properties node pulls the top 80 writable properties for Contacts, Companies, and Deals (excluding calculated/hidden/readOnly fields). The Filter Properties & Extract Attendees node drops internal emails matching INTERNAL_EMAIL_DOMAINS and keeps only external participants for further processing.

5. Search contact and look up deal associations

   For each external participant, the Search HubSpot Contact node searches by email, and the Lookup Contact Deal Associations node fetches the first associated deal. If the contact does not exist, the flow branches per UNKNOWN_CONTACT_ACTION (skip/notify/create).

6. Gemini AI analysis and structured extraction

   The Prepare AI Context node assembles the transcript, HubSpot property lists, and current contact/company/deal context. The AI Meeting Analysis node calls Gemini with a system prompt requiring: a Markdown summary, sentiment (positive/neutral/negative), intent (low/medium/high), an optional follow-up task, and contact_updates/company_updates/deal_updates whose keys must come from the provided property lists—no invented property names allowed.

   Execution prompt · Source text。AI Meeting Analysis node systemPromptTemplate field

   Copy prompt

   You extract CRM facts from a sales or customer meeting transcript and map them onto the HubSpot property schemas provided in the input.
   
   Rules:
   1. Produce summary_markdown as a clear Markdown meeting summary that includes summary, sentiment, intent, and actionable next steps.
   2. sentiment must be one of: positive, neutral, negative.
   3. intent must be one of: low, medium, high.
   4. Map extracted insights ONLY to property names that appear in the provided HubSpot Contact, Company, and Deal property lists. Never invent property names.
   5. contact_updates, company_updates, and deal_updates must be objects whose keys are exact HubSpot property names and whose values are extracted strings/numbers. Use {} when nothing maps.
   6. Never infer facts that are not supported by the transcript. Use empty objects/strings/false when information was not discussed.
   7. task.create_task should be true only when there is a concrete follow-up with a title.

   Setup guidance · Source text。AI Meeting Analysis node jsonSchemaExample field (output format reference)

   Copy prompt

   {"summary_markdown": "## Summary\nCustomer is evaluating a CRM migration.\n\n### Intent\nhigh\n\n### Sentiment\npositive\n\n### Next Steps\n- send proposal", "sentiment": "positive", "intent": "high", "task": {"create_task": true, "title": "Send implementation proposal", "due_date": "2026-09-02", "notes": "Include WhatsApp integration scope"}, "contact_updates": {"last_ai_meeting_summary": "Evaluating CRM migration"}, "company_updates": {"last_meeting_requirements": "WhatsApp integration"}, "deal_updates": {"last_meeting_budget": "50000 USD"}}

7. Execute HubSpot CRM sync

   The Process AI & Build CRM Sync node parses Gemini output, validates against the property allow-list, and builds a decision object (update_contact / update_company / update_deal / create_note / create_task / send_slack). The Execute HubSpot Sync node calls the HubSpot v3 API: batch upsert contacts (keyed by email), PATCH companies and deals, and create Note and Task objects associated to the relevant records. If the contact does not exist and policy is create, a new contact is created first.

8. Slack notifications and error monitoring

   If SEND_SLACK_ALERT is enabled and intent=high, or if an unknown contact triggers notify, the Optional Slack Notification node sends a formatted message to SLACK_CHANNEL. The Error Trigger catches workflow execution failures, and the Slack Execution Error node sends an alert with Execution ID, error node name, and error details.

   Execution prompt · Source text。Slack Execution Error node text field

   Copy prompt

   =🚨 *n8n Workflow Execution Failure*
   • *Workflow*: Sync Fireflies meeting notes to HubSpot CRM with AI
   • *Execution ID*: `{{ $json.execution.id || $execution.id }}`
   • *Error Node*: `{{ $json.execution.error.node.name || $json.execution.lastNodeExecuted || "unknown" }}`
   • *Error Message*: `{{ $json.execution.error.message || $json.message || "Unknown error" }}`

   Execution prompt · Source text。Optional Slack Notification node text field

   Copy prompt

   ={{ $('Route CRM Output').item.json.decision.notify_message }}

You are setting up an n8n workflow that automatically syncs Fireflies meeting notes into HubSpot CRM. Please confirm the prerequisites first: 1) You have API credentials for Fireflies, HubSpot, Google Gemini, and Slack; 2) You have filled INTERNAL_EMAIL_DOMAINS, SLACK_CHANNEL, and AI_MODEL in the Workflow Configuration node; 3) You have configured the Fireflies webhook to point to this workflow's fireflies-meeting-completed path. After building, validate with a test meeting: check that Validate & Normalize Meeting outputs correct attendees, that AI Meeting Analysis returns JSON with summary_markdown and intent, and that the corresponding Note appears in HubSpot.