You get:
- empty activity logs (no visibility into customer conversations)
- vague notes (“had a good call” — what does that mean?)
- missing next steps (deal stalls because no one knows what’s due)
- no institutional memory (rep leaves, knowledge leaves with them)
- forecasting based on incomplete data (garbage in, garbage out)
But structured summaries capture value:
- key topics: what was discussed?
- decisions: what was agreed?
- action items: who does what by when?
- sentiment: how did the customer feel?
- next steps: what happens next?
Without summarization, customer intelligence dies in transcripts.
This prompt transforms raw conversation into structured CRM logs.
Assume the role of a sales operations analyst who structures activity logs. Your task is to convert raw conversation into a structured CRM activity summary. Generate: 1. ACTIVITY METADATA - Type: [Call / Meeting / Email / Demo / Negotiation] - Date & Duration: [timestamp] - Participants: [list with roles] 2. KEY TOPICS COVERED - Topic 1: [summary of discussion] - Topic 2: [summary of discussion] - Topic 3: [summary of discussion] 3. CUSTOMER SENTIMENT - Overall: [Positive / Neutral / Negative / Mixed] - Evidence: [quote or observation] - Warning signs: [budget concerns, competitor mention, timeline slip] 4. DECISIONS MADE - Decision 1: [what was agreed] - Decision 2: [what was agreed] 5. ACTION ITEMS (with owners and due dates) | Action | Owner | Due Date | Priority | |--------|-------|----------|----------| | [task] | [person] | [date] | High/Med/Low | 6. CUSTOMER QUESTIONS - Question 1: [question asked] - Question 2: [question asked] 7. NEXT STEPS - What happens next - When the next activity is scheduled 8. LOG READY FOR CRM - A 2-3 sentence summary suitable for CRM notes field INPUTS: Conversation transcript or notes: [PASTE CALL TRANSCRIPT, MEETING NOTES, OR EMAIL THREAD] Activity type: [CALL / MEETING / EMAIL / DEMO / NEGOTIATION] Participants (if known): [E.G., "John (AE), Sarah (Customer - Procurement Lead)"] Customer stage: [PROSPECT / ACTIVE NEGOTIATION / CUSTOMER / CHURN RISK] RULES: - Extract decisions explicitly — "we agreed" is valuable, "we discussed" is not - Action items without owners are not action items (flag as "needs owner") - Capture customer questions (unanswered questions are risks) - Note sentiment changes (e.g., "started positive, ended concerned") - Keep the final CRM log to 2-3 sentences (nobody reads long notes) - Flag any mention of competitors or budget as high-priority signals
- Extract decisions explicitly — “we agreed to X” is valuable; “we discussed X” is not.
- Action items without owners are not action items (flag as “needs owner”).
- Capture customer questions — unanswered questions are risks.
- Note sentiment changes — “started positive, ended concerned” is critical context.
- Keep the final CRM log to 2-3 sentences — nobody reads long notes.
- Flag any mention of competitors or budget as high-priority signals for CRM-02.
Conversation transcript:
“John: Thanks for the demo. I think the product fits our needs. Sarah: Great, what about pricing? John: We need to stay under $50k. We’re looking at Competitor X but they’re more expensive. Sarah: I can send you a proposal by Friday. Also, can you connect me with your security team? John: Yes, I’ll set that up. Let’s talk next Tuesday to review.”
Activity type:
“CALL”
Participants:
“John (AE), Sarah (Customer – Procurement Lead)”
Customer stage:
“ACTIVE NEGOTIATION”
This framework improves outcomes by forcing:
- key topic extraction (what was actually discussed?)
- sentiment analysis (how did the customer feel?)
- decision capture (what was agreed?)
- action item extraction (who does what by when?)
- next step specification (what happens next?)
Failure modes this prevents:
- Vague notes — “good call” tells you nothing
- Missing action items — deals stall because no one knows what’s due
- Lost institutional knowledge — rep leaves, context leaves
- Unanswered questions — customer asked, no follow-up
This improves on: Free-text CRM notes that are never read. Structured summaries are searchable, reportable, and actionable.
Related to: CRM-04 (Intent Tagger) for classification; CRM-02 (Health Scanner) for risk signals.
Build Better AI Systems
Subscribe for advanced prompt engineering, AI coding tools, debugging frameworks, and practical strategies for developers and engineers.