Meeting Agent — AI Meeting Intelligence for Microsoft Teams
A consulting client needed a thinking partner in the room — not a recording tool, but something that listens live, coaches in real time, and handles all the documentation automatically when the meeting ends. We built it on Microsoft Teams using Azure OpenAI.
The situation
Consulting work lives in meetings. Note-taking steals attention. Follow-up documents take hours after an already full day. Action items get missed or assigned to no one. The best question often arrives ten minutes too late.
What we built
- AI agent joins every Teams meeting as a participant and reads the live transcript continuously
- Every ten lines, a fast AI model classifies what's happening — unanswered questions, uncertainty, agenda drift, commitments without owners
- CC mode: tips appear in Teams chat, visible to all. BCC mode: private DM to the consultant only
- Six AI personas tuned to different meeting types: Sales, Technical, Analyst, Interviewer, General, Custom
- Knowledge base with RAG — upload playbooks, pricing sheets, client briefs; relevant passages surfaced live during the call
- Post-meeting pipeline: AI summary and action items posted to Teams chat and archived to SharePoint automatically
- Recording and transcript retrieved from Microsoft's infrastructure and filed — retrying for up to 45 minutes
What changed
Every meeting produces consistent, structured outputs without anyone spending time after the fact to create them. Consultants stay fully present. Action items surface with owners before anyone has left the call.
Client kept anonymous at their request.