How to Prompt an AI Note-Taking & Conversation Assistant

September 10, 2025 · 7 min read

Why Prompting Matters

An AI conversation assistant like the WIN System can transcribe hours of meetings and analyze them in seconds. But the quality of the output depends entirely on how you ask. A vague prompt like "summarize this" produces a generic paragraph. A precise prompt produces structured, actionable intelligence.

This guide covers the prompting strategies that turn a basic transcript summary into a powerful productivity tool.

The Anatomy of a Good Prompt

Every strong prompt has three elements:

1. Role — Tell the AI who it is. "You are a sales meeting analyst" produces very different output from "You are a project manager."
2. Task — Be specific about what you want. "Extract all action items with owners and deadlines" is far better than "summarize."
3. Format — Tell the AI how to structure the response. Bullet points? JSON? A table? If you don't specify, you get walls of text.

Prompt Templates for Common Scenarios

1. Meeting Summary with Action Items

Analyze this meeting transcript. Provide: (1) A 3-sentence executive summary, (2) A bullet list of every decision made, (3) A bullet list of every action item with the responsible person and deadline if mentioned, (4) Any unresolved questions or open issues.

2. Sales Call Intelligence

You are a sales analyst. Review this transcript of a sales call. Extract: the prospect's company name, their main pain point, any competitors mentioned, their budget or timeline if stated, and the agreed next step. Format as a structured list.

3. Technical Discussion Breakdown

Analyze this technical discussion. List every technical requirement mentioned, any architecture decisions made, blockers or risks identified, and any follow-up tasks. Use bullet points grouped by category.

4. Training / Lecture Notes

You are a student taking detailed notes. From this lecture transcript, extract the key concepts taught, any definitions given, examples used to illustrate points, and any homework or reading assignments. Organize by topic heading.

5. Customer Support Triage

Analyze this support call transcript. Identify: the customer's issue, the steps already attempted, the resolution (if reached), and any follow-up required. Also rate the customer's sentiment as positive, neutral, or frustrated.

Advanced Prompting Techniques

Use the Screenshot

WIN System sends a screenshot along with your transcript. You can reference it directly in your prompt:

Look at the screenshot and the transcript together. The screenshot shows the presentation slide being discussed. What is the key metric shown on the slide, and how does the speaker's commentary relate to it?

Chain Prompts Across Sessions

After the first "Ask the AI" call, you can copy the response, paste it into the prompt field, and ask a follow-up:

Based on the summary you just generated, draft a follow-up email to the client thanking them for the call and confirming the three action items.

Combine with RAG

If you've uploaded documents to WIN System's RAG feature, the AI automatically cross-references them. Your prompt can explicitly invoke this:

The customer mentioned "Enterprise Plan pricing." Check my uploaded pricing documents and compare what the customer was told on this call against our current published pricing.

Common Prompting Mistakes

• "Summarize this" — Too vague. Always specify what kind of summary, how long, and what to include.
• Not specifying format — If you want bullet points, say "bullet points." If you want JSON for CRM import, say "JSON."
• Ignoring the screenshot — The screenshot is part of the context. Use it, especially for presentations and screen shares.
• Treating the AI as a search engine — The AI has the transcript right in front of it. Don't ask "what is RAG?" — ask "what did the speaker say about RAG in this meeting?"

Conclusion

Prompting is a skill, and like any skill it improves with practice. Start with the templates above, modify them for your workflow, and pay attention to which prompts consistently give you the best results. The more specific you are about role, task, and format, the more useful the AI's output becomes.