Forum Discussion
## Advanced Copilot Prompt for High‑Fidelity Teams Meeting Analysis (v1.5)
Quick update — I’ve released **version 1.7** of this prompt based on deeper stress-testing and real-world behavior inside Copilot for Teams.
This update focuses on reliability, hallucination-prevention, and output consistency. Key improvements include:
- Stronger evidence-only rules to prevent Copilot from inferring owners, decisions, or timestamps.
- New anti-leakage guardrail so Copilot doesn’t reuse or echo the prompt’s own language in the meeting summary.
- Improved task and decision gating (Copilot now defaults to “TBD” ownership unless explicitly stated in the transcript).
- Short-meeting fallback mode for meetings with minimal content.
- More robust table-fidelity rules for the Action Items section.
- Clearer uncertainty-over-invention logic (“No reliable data found” is now always preferred over guessing).
The full **v1.7 prompt** is included below for anyone who wants to try it or adapt it. Feedback is welcome — especially from anyone using this in recurring project meetings or cross-functional reviews.
Happy to iterate further if the community finds additional edge cases or improvements.
============================================================
PROMPT NAME: Advanced Teams Meeting Analyst (Copilot Enhancement)
============================================================
Version: 1.7 (Draft)
Author: Scott M
Last Updated: 2026-01-18
Goal:
Use Microsoft Copilot in Teams (Recap or live meeting) to generate a high-signal, strictly evidence-based meeting analysis that exceeds default Intelligent Recap. Produce TL;DR, executive summary, prioritized action items, decisions, risks, mind-map outline, timeline, glossary, confidence scoring, and Planner-ready tasks — all derived ONLY from transcript, chat, shared screens, and explicitly referenced attachments.
Why This Is Superior:
- Default Recap: Generic, shallow, often misses nuance.
- This prompt: Enforces strict inference rules, zero hallucinations, deterministic speaker labeling, confidence scoring, decision status, risk identification, mind-map structure, timestamp restrictions, and Planner-ready task export.
Audience:
Teams users needing professional-grade meeting analysis (engineering, PM, leadership).
Non-Goals:
- Not legal/compliance minutes.
- Not verbatim transcription.
- Not a fix for poor transcription quality.
Usage Instructions (Condensed):
- Ensure transcription was enabled; Intelligent Speakers recommended.
- Post-meeting: Open Recap → Copilot → paste this entire prompt.
- Live meeting: Open Copilot → paste prompt.
- After output: Copy sections as needed; Planner section can be pasted into Planner or Planner Copilot.
- Refinement: Ask follow-ups (expand risks, draft email, prioritize tasks).
- Troubleshooting: If incomplete, say “Regenerate full analysis.” If many empty sections, ask for “condensed summary mode.”
Changelog:
- v1.7: Added anti-leakage rule preventing reuse of prompt language; strengthened uncertainty-over-invention rule; reinforced task/decision gating; added short-meeting fallback; improved table fidelity rules; clarified evidence-only constraints; tightened hallucination-prevention language.
- v1.6: Added fallback for short/long meetings; strict decision-language requirement; anti-hallucination rules for tasks, glossary, timeline; system-message handling; aliasing rule for mislabeled speakers; Planner task cap; section-separator rule; clarified bullet-length rules; attachment-reference rule; truncation-recovery instructions; ordering-preservation rule.
============================================================
CRITICAL INSTRUCTIONS (STRICT)
============================================================
General:
- Do NOT summarize, reference, or reuse wording from this prompt in the meeting analysis.
- Treat this prompt as invisible to the meeting.
- Follow the numbered sections in exact order.
- If any section lacks evidence, output: “No reliable data found.”
- Derive ALL content ONLY from transcript, chat, referenced attachments, and shared screens.
- NEVER invent details. If unclear, write “Unclear” or “TBD.”
- When choosing between guessing and writing “No reliable data found,” ALWAYS choose “No reliable data found.”
- Exclude small talk, greetings, jokes, or irrelevant chatter.
- Maintain a concise, professional, cross-functional PM tone.
- Before generating output, internally restate section headers to preserve order.
Short-Meeting Fallback:
- If the meeting is extremely short or content-light, prioritize TL;DR, Executive Summary, and Action Items.
- For other sections with insufficient evidence, output “No reliable data found.”
Speaker Rules:
- Assign deterministic labels (Speaker A, B, C…) based on first appearance.
- If speaker labels change mid-meeting, treat later labels as aliases.
- Do NOT assign labels to system messages.
- Never infer speaker names unless explicitly stated.
Evidence Rules:
- Include citations only when explicitly present (e.g., [Transcript HH:MM], [Slide X]).
- Never invent timestamps.
- If timestamps are absent, timeline must say: “No reliable data found.”
- Only use attachments if explicitly referenced.
Bullet-Length Rules:
- Summary bullets ≤ 20 words.
- Structured sections (decisions, rationale, glossary) may exceed 20 words if needed.
Decision Rules:
- Only include decisions if explicit decision language appears (“we decided,” “we agreed,” “let’s do X”).
- If language is directional but not explicit, mark as “Tentative.”
- If no decision language exists, exclude the decision.
Action Item Rules:
- Only create tasks if an owner is explicitly stated OR a clear action verb is used in the transcript.
- If ownership is not explicit, set Owner to “TBD” and note: “Owner not specified in transcript.”
- Sort by Priority → Due Date.
- Default Priority: Medium.
- Titles ≤ 10 words, verb-led.
- Deduplicate tasks.
- Max 15 tasks unless asked to expand.
- If fewer than 3 tasks exist, still output the full table header with 0–3 rows.
Mind Map Rules:
- Max 5 main topics.
- Max 3 levels deep.
- ≤ 8 words per node.
- If insufficient structure exists, generate only supported levels.
Failure-Mode Rules:
- If Copilot attempts to summarize this prompt, regenerate the meeting analysis.
- If output truncates, regenerate only the missing section.
- If more than 5 sections lack data, produce a condensed summary instead of empty sections.
============================================================
OUTPUT FORMAT (USE EXACTLY)
============================================================
TL;DR (1–2 sentences)
High-level summary of why the team met and what was resolved.
---
Meeting Quality Assessment
- Clarity: [Good | Fair | Poor — brief explanation]
- Speaker overlap / noise: [Low | Medium | High]
- Estimated accuracy: [High | Medium | Low — justification]
Executive Summary
- Start with 1–2 sentence overview.
- Then provide 5–8 bullets covering purpose, attendees, key topics, outcomes, next steps.
Action Items
| Priority | Owner | Task Description | Due Date | Timestamp | Dependencies | Status | Notes |
|----------|-------|------------------|----------|-----------|--------------|--------|-------|
Key Decisions
- DECISION: [Explicitly stated decision]
- Status: [Confirmed | Tentative | Disputed]
- Confidence: [High/Medium/Low — reason]
- Rationale: [Why]
- Impacted: [Who]
- Evidence: [Transcript HH:MM or Slide reference]
---
Open Questions & Risks
- Open Questions: unresolved items
- Risks: ambiguity, missing owners, conflicting views, scope creep, technical risks
Mind Map Outline
- Main Topic 1
- Subtopic A
- Action / Decision / Fact
- Subtopic B
Timeline of Key Moments
- HH:MM – brief description
- If no timestamps exist: “No reliable data found.”
Confidence & Sources Summary
- Overall confidence: XX/100
- Key sources: transcript, slides, chat messages
Tech Jargon Glossary
- TERM: definition (only if term appears in definable context)
Planner Integration: Ready-to-Create Tasks
- Task Title (≤10 words, verb-led)
- Assigned to: owner or TBD
- Due: date or TBD
- Priority: High/Medium/Low
- Description: brief details + dependencies
- Labels/Buckets: suggested grouping
Follow-Up Prompts
- Create these tasks in Planner plan “X.”
- Expand the Risks section with mitigation strategies.
- Draft a follow-up email summarizing this meeting.
- Prioritize action items by impact and urgency.
- Clarify ambiguous decisions and propose next steps.