Your organization already said the answer. It was in a meeting nobody can find.
Meeting Intelligence Assistant takes Microsoft Teams transcripts, strips confidential content, passes them through Azure AI Content Safety, then chunks and indexes them in Azure AI Search with date and participant metadata. Each user searches only the meetings they actually attended — and the whole pipeline is triggered automatically the moment a meeting ends.
Meeting Intelligence Assistant — Meeting Intelligence & RAG Assistant
The problem underneath.
Transcription is solved. Retrieval is not. Most organizations now generate transcripts automatically and then do nothing with them, because a folder of 4,000 transcripts is less useful than no transcripts at all — you cannot find the decision, so you ask again in the next meeting.
The hard part is not summarisation, it is permission-aware retrieval. A meeting transcript is one of the most sensitive documents an organization produces: it contains half-formed opinions, salary discussions, deal terms and things said in confidence. A search index that returns any of that to any employee is not a productivity tool, it is an incident.
This product's central design decision solves exactly that. The Assistant on Foundry runs its RAG search against index metadata for meeting date and participants, so a user can only ever retrieve from meetings they were in. Combined with the privacy-stripping and Content Safety stages before indexing, the result is a searchable institutional memory that does not create a disclosure risk.
Input, decision, output.
The full path a request takes through Meeting Intelligence Assistant — nothing hidden in the middle.
Teams transcripts
Meeting transcripts taken as the source and fed into the processing pipeline.
Privacy strip & Content Safety
Confidential content is removed and text is passed through Azure AI Content Safety.
Azure AI Search index
Content chunked and indexed with date and participant metadata.
Foundry Assistant RAG
Grounded retrieval scoped by the user's own meeting history.
Graph API automation
The pipeline starts by itself the moment the meeting ends.
What it actually does.
Automatic ingestion
Microsoft Graph API triggers the whole process when a meeting ends — no manual step, no forgotten upload.
Privacy stripping
Confidential content is removed before anything is indexed, not filtered at query time.
Content Safety gate
Text passes through Azure AI Content Safety stages before entering the index.
Participant-scoped access
Each user searches only within meetings they attended; the participant list is the permission boundary.
Grounded answers
Responses are grounded on indexed content with date and participant metadata, not generated from memory.
Azure-native deployment
Runs end to end inside your Azure tenancy — Teams, Content Safety, AI Search and Foundry.
Who this is for — and who it isn't.
Most vendors only answer the first half. The second half saves everybody a quarter.
A good fit if
- You run on Microsoft 365 and Teams meetings are where decisions actually get made
- You already generate transcripts and are doing nothing useful with them
- You have the Azure footprint to host this inside your own tenancy
- Institutional memory loss is a real, named cost in your organization
Probably not if
- Your organization is not on Microsoft 365 — the Graph API trigger is central to the design
- You want a cross-organization knowledge base where everyone can search everything
- Meetings are not where your decisions are recorded
Questions we get asked.
Can someone search a meeting they were not in?
No. Participant metadata is part of the index and forms the retrieval boundary. This is enforced at query construction, not as a post-filter, so there is no path by which content from an unattended meeting surfaces in an answer.
What does privacy stripping actually remove?
Content classified as confidential is removed before indexing, and the remaining text passes through Azure AI Content Safety stages. The exact classification rules are configured to your policy — what counts as confidential differs between a bank and a manufacturer.
Does anything leave our Azure tenancy?
No. The pipeline is Azure-native by design: Teams, Content Safety, AI Search and Foundry all run inside your own subscription.
What if a meeting has external participants?
External participants are part of the participant list and therefore part of the permission model. Whether externally attended meetings are indexed at all is a policy decision made at configuration.
How is this different from Copilot meeting recap?
Recap summarises one meeting for its attendees. This builds a persistent, permission-scoped index across all of a user's meetings, so the question "what did we decide about X last quarter" has an answer — grounded, cited and access-controlled.
Often deployed alongside.
Sales Mentor
Sales Mentor is a coach agent that talks — by voice — with sales and marketing teams before and after meetings and activities.
AI-BUS Agent Monitoring
AI-BUS Agent Monitoring is a single pane of glass for subscribing to and observing every AI agent running across your organization.
Dispatcher Agent
Dispatcher Agent is an autonomous AI agent that monitors your company's dedicated email addresses, reads both the message and its attachments, decides which business process it belongs to, and starts that process in your CRM or ERP — or completes the action directly.
See Meeting Intelligence Assistant on your own data.
Start with a free 90-minute assessment. We map where this fits in your organization, what it would touch, and what it would return — then you decide.