7.1 Using Copilot for Meetings: Setup & Live Intelligence

7.1 Using Copilot for Meetings: Setup & Live Intelligence

Quick Answer: Microsoft 365 Copilot in Microsoft Teams operates in two primary modes: With Transcription (transcription or recording enabled, delivering persistent records, speaker attribution, and post-meeting Intelligent Recap) and Without Transcription / "Only during the meeting" (ephemeral, in-memory speech-to-text processing where no transcript or audio is retained once the call ends). During live meetings, Copilot answers real-time queries ("What did I miss?", "What are the key arguments for each proposal?", "What questions remain unresolved?"). Post-meeting, Intelligent Recap delivers AI-generated notes, topic chapters, speaker timeline markers, decision logs, and structured action item matrices with assigned owners.

Meetings represent the core collaborative engine of modern enterprise work—yet they are also a primary source of cognitive fatigue, lost context, and administrative drag. Professionals routinely juggle attending overlapping discussions, capturing accurate meeting minutes, tracking assigned deliverables, and onboarding late participants. Microsoft 365 Copilot in Teams fundamentally transforms meeting productivity by acting as an intelligent, real-time participant and automated scribe.

To leverage Copilot effectively and pass the AB-730 exam, business professionals must master the operational mechanics of real-time in-meeting queries, understand the compliance and architectural differences between transcription modes, and utilize post-meeting Intelligent Recap features for executive governance.

---

1. Operating Modes: With Transcription vs. Without Transcription

One of the most critical architectural distinctions in Microsoft 365 Copilot is how Teams handles meeting audio, transcription, and data persistence. Meeting organizers can configure Copilot to run either with full transcription or in an ephemeral mode.

┌─────────────────────────────────────────────────────────────────────────────┐
│                  TEAMS COPILOT MEETING OPERATING MODES                      │
├──────────────────────────────────────┬──────────────────────────────────────┤
│         WITH TRANSCRIPTION           │       WITHOUT TRANSCRIPTION          │
│    (Persistent Grounding & Recap)    │    ("Only During the Meeting")       │
├──────────────────────────────────────┼──────────────────────────────────────┤
│  • Audio converted to text via       │  • Audio processed in-memory via     │
│    Azure Speech Services             │    ephemeral speech-to-text pipeline │
│  • Persistent transcript stored in   │  • Zero audio or transcript data     │
│    Exchange Online / OneDrive        │    stored at rest                    │
│  • Full speaker diarization &        │  • Real-time speaker attribution     │
│    attribution retained              │    available only while call is live │
│  • Real-time in-meeting Q&A active   │  • Real-time in-meeting Q&A active   │
│  • Post-Meeting Intelligent Recap,   │  • NO post-meeting recap, transcript,│
│    topic chapters & follow-ups       │    or post-call Copilot queries      │
└──────────────────────────────────────┴──────────────────────────────────────┘

Mode 1: With Transcription (Standard / Persistent Mode)

In this mode, live transcription or recording is initiated (either automatically by policy or manually by an authorized participant).

• Processing & Storage: Spoken audio is processed by Azure Speech Services, generating a synchronized text transcript with speaker diarization (identifying who spoke what). The transcript file is stored persistently in Exchange Online and linked to the meeting recording in OneDrive for Business / SharePoint.
• Capabilities: Participants can query Copilot both during the live meeting and indefinitely after the meeting concludes. It powers the full Intelligent Recap experience, including AI notes, automated chapters, follow-up recommendations, and transcript search.
• Best For: Standard business reviews, recurring sprint syncs, project kickoffs, and cross-functional planning sessions where an auditable record and asynchronous review are required.

Mode 2: Without Transcription ("Only During the Meeting")

For sensitive, legally restricted, or highly confidential discussions, organizations may prohibit persistent meeting recordings and transcripts. Microsoft 365 Copilot supports an ephemeral operating mode configured via the meeting option: "Allow Copilot: Only during the meeting."

• In-Memory Ephemeral Processing: Spoken audio is converted into text in temporary RAM buffers to process user prompts during the call. The AI analyzes the ephemeral transcript chunk, streams the response into the Copilot side pane, and immediately discards the raw audio and transcript from memory.
• Zero Data Retention: Once the meeting concludes, no transcript, recording, or Copilot interaction history exists. The Recap tab in Teams will display meeting metadata (attendees, duration, shared files) but will state that AI recap and transcript are unavailable.
• Compliance & Legal Use Cases: Ideal for confidential HR hearings, executive compensation reviews, sensitive M&A negotiations, attorney-client privileged consultations, and works council-regulated regions where employee voice recording is legally restricted.

Architectural Comparison Matrix

Exam Warning: On Exam AB-730, pay close attention to scenario requirements regarding data retention. If a question describes a legal department requiring real-time AI assistance during settlement negotiations without creating an auditable transcript subject to future legal discovery, the correct configuration is "Allow Copilot: Only during the meeting".

---

2. Real-Time In-Meeting Queries & Live Intelligence

During an active Teams meeting, participants with a Copilot license can open the Copilot pane from the meeting toolbar to query the live discussion. Copilot processes the conversation in real time, synthesizing complex spoken dialogue into concise, actionable insights without interrupting the flow of the meeting.

┌─────────────────────────────────────────────────────────────────────────────┐
│                     LIVE IN-MEETING COPILOT QUERIES                         │
├─────────────────────────────────────────────────────────────────────────────┤
│  [1. CATCH-UP & CONTEXT TRIAGE]                                             │
│   • "What did I miss so far?"                                               │
│   • "Summarize the last 15 minutes of discussion."                          │
│                                                                             │
│  [2. ARGUMENT & PROPOSAL ANALYSIS]                                          │
│   • "What are the main arguments for and against the proposed vendor?"      │
│   • "Create a table comparing Sarah's proposal with David's alternative."  │
│                                                                             │
│  [3. CONSENSUS & UNRESOLVED ISSUES]                                         │
│   • "Where do the participants agree, and where is there disagreement?"     │
│   • "What questions were raised that have not yet been answered?"           │
│                                                                             │
│  [4. SENTIMENT & PERSPECTIVE TRACKING]                                      │
│   • "Who expressed concerns about the Q4 launch timeline, and why?"         │
│   • "What feedback did the Finance Director provide regarding the budget?"  │
└─────────────────────────────────────────────────────────────────────────────┘

Core In-Meeting Query Patterns

1. Late Joiner Catch-Up: When joining a meeting 20 minutes late, a user can prompt Copilot: "Summarize what was discussed before I joined, highlighting any decisions made." Copilot returns a bulleted summary of prior dialogue without requiring the user to interrupt the speaker or ask colleagues in chat.
2. Tracking Debates and Counterarguments: In multi-stakeholder debates, participants often talk past one another. Prompting: "What are the key points of disagreement regarding the hybrid cloud migration?" extracts the distinct positions held by different speakers.
3. Unresolved Question Discovery: Before concluding a meeting, the facilitator can prompt: "List all questions asked during the meeting that remain unanswered." Copilot scans the dialogue for queries that lacked follow-up answers, preventing critical operational gaps.
4. Structured Pros/Cons Tables: Users can request on-the-fly comparative matrices: "Generate a table listing the pros, cons, and budget impact for each architectural option discussed."

---

3. Meeting Voice Copilot vs. Meeting Chat Copilot

A common area of confusion in Microsoft Teams is the distinction between spoken audio and side-channel text chat. Teams meetings typically generate two parallel streams of communication:

┌─────────────────────────────────────────────────────────────────────────────┐
│                   VOICE COPILOT VS. CHAT COPILOT IN TEAMS                   │
├──────────────────────────────────────┬──────────────────────────────────────┤
│       MEETING VOICE COPILOT          │        MEETING CHAT COPILOT          │
│                                      │                                      │
│  • Source: Spoken audio stream       │  • Source: Written text messages,    │
│    captured via microphone           │    links, and files in meeting chat  │
│  • Processed via: Speech-to-text     │  • Processed via: Standard Teams     │
│    transcript pipeline               │    chat message Graph index          │
│  • Activated via: In-meeting Copilot │  • Activated via: Chat pane Copilot  │
│    toolbar button                    │    or main Teams chat interface      │
│  • Query focus: Spoken arguments,    │  • Query focus: Shared URLs, text    │
│    verbal commitments, tone, debates │    comments, asynchronous reactions  │
└──────────────────────────────────────┴──────────────────────────────────────┘

Interacting with Both Modalities

• Querying the Voice Stream: Asking "What did Alex say about the Q3 budget?" queries the spoken transcript captured by Azure Speech Services.
• Querying the Written Chat: Asking "Summarize the links and resources shared in the meeting chat" queries the persistent text chat repository.
• Synthesized Cross-Modality Queries: When a user opens Copilot in the meeting window, Copilot prioritizes the spoken transcript but can cross-reference chat contributions if explicitly instructed (e.g., "Did anyone in the chat agree with Maria's verbal proposal?").

---

4. Prompt Engineering for Teams Meetings: GCSE in Action

To extract the highest value from Copilot during and after meetings, business professionals must structure their prompts using the GCSE Framework (Goal, Context, Source, Expectations):

---

5. Realistic Business Scenario: High-Stakes Multi-Vendor Architecture Sync

Business Context

Marcus is a Principal Technical Program Manager at Contoso Financial. He leads a 90-minute monthly architecture synchronization call with 25 participants, including internal security architects, infrastructure leads, and two third-party software vendors (Fabrikam and Northwind).

Operational Challenges

1. Marcus must actively facilitate the complex technical debate without spending the entire call manually typing minutes.
2. Several executive leaders joined the call 25 minutes late due to conflicting meetings.
3. The team debated three competing database migration strategies, with contentious disagreements on downtime tolerances.

┌─────────────────────────────────────────────────────────────────────────────┐
│                     CONTOSO ARCHITECTURE SYNC WORKFLOW                      │
│                                                                             │
│   [LIVE CALL: 90 MIN]                                                       │
│    • Transcription enabled by Marcus (Organizer)                            │
│    • 00:25: Late VP queries: "What did I miss regarding cloud storage?"     │
│    • 00:55: Marcus queries: "Compare Fabrikam vs Northwind proposals"       │
│    • 01:25: Facilitator queries: "List all unanswered security questions"   │
│                                                                             │
│   [POST-CALL: INTELLIGENT RECAP]                                            │
│    • Auto-generated Topic Chapters: Storage (05:10), DB Strategy (32:40)    │
│    • Marcus exports Action Item Table to Microsoft Planner                  │
│    • Copilot drafts 1-page executive memo from transcript in Word           │
└─────────────────────────────────────────────────────────────────────────────┘

Execution & Results

• During the Call (Minute 25): The late-joining VP opens Copilot and submits: "Summarize the discussion on cloud storage tiers and state if any budget decisions were made." Within seconds, the VP is fully aligned without interrupting Marcus.
• During the Call (Minute 55): To resolve a circular debate, Marcus prompts: "Create a side-by-side comparison table of Fabrikam's API approach vs Northwind's microservices approach, listing cost, estimated latency, and downtime risks." Marcus shares the generated table in meeting chat, driving immediate group consensus.
• Post-Call Governance: Marcus navigates to the Recap tab. Copilot has parsed 14 discrete action items, correctly attributing task ownership to Fabrikam's lead engineer, Northwind's architect, and Contoso's security manager. Marcus copies the structured table directly into Microsoft Planner for sprint tracking.

---

6. Exam Tips & Governance Warnings

[!IMPORTANT] Critical Knowledge for Exam AB-730

1. Transcript Prerequisite: Full post-meeting Intelligent Recap features (topic chapters, speaker timeline markers, and persistent post-call chat) require transcription or recording to be turned on. If the meeting operates in "Only during the meeting" mode, no post-meeting recap is generated.
2. Tenant Policy Controls: In the Teams Admin Center (TAC), administrators control Copilot availability via Meeting Policies > Allow Copilot. The options are: Enabled, EnabledWithTranscript, EnabledUserOverride, or Disabled.
3. Explicit Recording Consent: When tenant compliance enforces explicit consent, all participants must agree to recording/transcription before unmuting or sharing video.
4. Guest & External Access: External guest participants can only interact with Copilot during the meeting if tenant cross-tenant policies and meeting settings explicitly permit external transcript access.