Skip to main content

​
Meeting Intelligence Pipeline

​
Overview

The Meeting Intelligence Pipeline turns your project meetings into structured insights and action items automatically. When you record meetings with tl;dv, the platform processes transcripts, extracts decisions and commitments, files everything to your project vault, and distributes action items to owners via Slack, WhatsApp, or email. Your AI agents - SupaSherpa and the AI Product Manager - gain meeting context to provide better advice and catch risks before they escalate.

​
Step-by-Step Guide

1. Record your meeting with tl;dv Use tl;dv to record your next project meeting on Zoom, Google Meet, or Microsoft Teams. The platform automatically receives a notification when the recording completes. 2. Wait for automatic processing Within 2 minutes, the system processes your transcript and generates structured meeting minutes. You receive a notification when processing completes. Minutes include:
  • Meeting title, date, duration, and attendees
  • Discussion summary (3-5 bullet points per major topic)
  • Decisions made during the meeting
  • Commitments with owners and deadlines
3. Review action items Each attendee receives their action items via their preferred channel (check your Channel Preferences at /privacypreferences). Items show up with the owner’s name, description, deadline (if mentioned), and a quote from the transcript showing where it was discussed. 4. Track action items on your dashboard Visit the SupaSherpa interface (/supasherpa) to see all your action items. SupaSherpa proactively mentions overdue items when you chat. Mark items as IN_PROGRESS or COMPLETED as you work through them. 5. Access meeting history Find all processed meetings in your project’s Data Room (/dataroomlist). Search by keyword to locate specific discussions or decisions. The AI Product Manager monitors meetings for risk indicators and alerts your sherpa if issues emerge. 6. Ask SupaSherpa about meetings When chatting with SupaSherpa, reference recent meetings. The VA retrieves meeting content and references specific decisions or action items to give contextual advice.

​
Common Questions

Q: What happens if I forget to record a meeting with tl;dv? A: Use the manual meeting creation endpoint to upload notes or a transcript after the fact. The system processes manual uploads the same way as automatic recordings. You lose real-time processing but gain the same structured output and action item distribution. Q: Can the system track action items for people not on the platform? A: Yes, but with limitations. If someone mentioned in the meeting isn’t registered, the action item is created with owner set to “unmatched.” You receive a notification to manually assign it to the correct person. Q: How long does it take to process a meeting? A: Processing completes within 120 seconds for meetings up to 60 minutes. Action items are distributed to owners within 60 seconds after processing. If your meeting is longer than 3 hours, the system processes the first 100,000 tokens and logs a warning about truncation. Q: What if the same meeting webhook arrives twice? A: The system deduplicates automatically using the tl;dv meeting ID. Duplicate webhooks return success but create no new processing job. You won’t see duplicate minutes or action items. Q: Can I prevent meeting notifications during off-hours? A: Yes. Action item distribution respects quiet hours - no delivery between 22:00-07:00 in your local time zone. Set your channel preferences at /privacypreferences.

​
Troubleshooting

Issue: Meeting processed but no action items extracted Check the meeting minutes in your Data Room. If the transcript contains vague commitments like “we should probably look into this,” the AI may not extract them. Action items require explicit statements like “Sarah will do X by Friday.” Review the transcript and manually create action items if needed. Issue: Meeting filed to the wrong project vault This happens when attendees are linked to multiple jobs. The system files to all relevant vaults (maximum 5). If a meeting appears in the wrong vault, remove it from the Data Room. To prevent this, ensure meeting invites include only people working on that specific project. Issue: SupaSherpa says “meeting still being processed” hours later Processing failed after 3 retry attempts. Check the meeting status in your Data Room. If status shows PROCESSING_FAILED, you receive a notification explaining the failure. Re-upload the transcript manually or contact support if the issue persists. Issue: Action items missing deadlines The AI extracts deadlines only when explicitly mentioned in the transcript. Phrases like “by next week” or “soon” don’t generate deadline fields. Review action items after processing and add deadlines manually via the API or by updating them in your project management tool. Data Room (/dataroomlist) - Meeting minutes and transcripts are stored here. Use the Data Room to search across all project documentation, including meeting history. The AI uses this content to answer questions and provide recommendations. SupaSherpa (/supasherpa) - Your AI virtual assistant gains meeting context from processed transcripts. Ask about recent meetings, get reminded of overdue action items, or request summaries of decisions made across multiple meetings. AI Product Manager Dashboard (/pipelinedashboard) - View risk signals and opportunities detected from meeting content. The PM agent monitors for timeline slippage, budget concerns, scope creep, and team conflicts mentioned during meetings, alerting your sherpa before problems escalate.