Build Your Core Asset

Initial Setup Checklist

  1. Obtain explicit permission and access to meeting recordings/transcripts.
  2. Use Prompt 1 to instruct the AI to extract specific categories (Decisions, Actions, Pending).
  3. Manually read the AI output and compare it against the transcript for accuracy.
  4. If the AI assigns a task but the owner is unclear in the transcript, manually flag it as "[Needs Clarification]".
  5. Deliver the final structured log to the client within the agreed timeframe (usually 24 hours).

Action Prompts

PROMPT 1: Structured Extraction
Act as an executive assistant. I will provide a meeting transcript. Extract the information into three sections: 1. Firm Decisions Made. 2. Action Items (Format: [Task] - [Owner] - [Deadline]). 3. Pending/Unresolved Questions. Do not invent owners or deadlines. If a detail is missing, write '[Not specified]'. Here is the transcript: [PASTE TRANSCRIPT]
PROMPT 2: Formatting for Handoff
Review the extracted notes. Format them into a professional email update. Use bolding for names and deadlines so they stand out. Keep the tone professional and concise.

Finished Example

This is an illustrative finished example for educational purposes.
A Slack message formatting: **Meeting Summary: Q3 Planning** **Decisions:** - We will launch the ad campaign on Sept 1st. - The budget is capped at $5,000. **Action Items:** - **Dave:** Draft ad copy (Due: Thursday) - **Sarah:** Approve final budget (Due: [Not specified - Please confirm]) **Parking Lot:** - Discuss the Q4 holiday promo next week.

Common Mistakes

  • Inventing decisions when the transcript is vague

    Impact: You assign a task to someone who never agreed to it, causing internal team conflict and making you look incompetent.

    Correction: Always flag uncertainty. If it wasn't explicitly decided, put it in the 'Pending' or 'Clarification Needed' section.

  • Violating confidentiality

    Impact: You leak sensitive company data by pasting it into public/unsecured AI models.

    Correction: Only use enterprise AI tiers with zero-data-retention policies, or explicitly clear the tool usage with the client first.

  • Delivering massive blocks of text

    Impact: The client still has to read a huge document, defeating the purpose of the service.

    Correction: Keep it scannable. Bullet points, bold names, and extreme brevity.

Failure Recovery Warning: Diagnostic guidelines for meeting note services.

Practice Execution

  1. Find a free transcript of a public meeting (e.g., a city council meeting or a public tech company Q&A).
  2. Run Prompt 1 on the transcript.
  3. Review the AI output carefully. Does it accurately reflect who is doing what?
  4. Manually edit the output to fix any AI hallucinations or missed nuances.
  5. Format it into a clean, 1-page summary document.
Supplementary Methodology & Evidence
Methodology & Evidence: Practice using public transcripts ensures you build the critical skill of identifying hallucinated action items before handling real, sensitive client data.

Ready to execute?

This playbook explains the work. My Mission is where you DO and TRACK the work.

Start Building