Mine Your Transcripts for Latent Gold With a Structured Extraction Prompt
The Insight
Most professionals record their conversations — coaching calls, mastermind sessions, podcast appearances, presentations — and then do nothing with them. The transcript sits in a folder. The value stays locked.
A single, well-constructed extraction prompt can turn any meeting transcript into a prioritized content brief. The trick is prompting the model not just to skim for explicit statements, but to read between the lines — catching the “half-said frustrations,” the “excited tangents,” and the pauses when someone is “onto gold.”
Donald Kihenja shared this prompt verbatim in the October 16 session chat.
The Prompt
Scan this meeting transcript for AI ideas useful to solo knowledge entrepreneurs — folks running their own show, juggling clients, tech, and dreams. Extract three to ten nuggets: practical tips, tools, or subtle themes that solve problems, spark hope, or ease fears. Don’t just skim — read like you’re one of them: catch the half-said frustrations, the excited tangents, that pause when someone’s onto gold. If they rant about time drains, hear “AI automation!” If they dream big, dig for their real craving — freedom, impact, less mess. Write each in first-person, collegial tone: warm, pro-but-friendly, like you’re swapping war stories over coffee. For emotional hits, weave in fables or pep talks — acknowledge fears (tech overload, imposter vibes) and flip to empowerment. Technical stuff? Keep it crisp, no fluff. Per nugget, list five content angles: newsletter hook, email essay outline, quick fable, tool demo, or fear-busting pep talk. Stay true — no made-up nonsense.
What Makes This Prompt Work
Several deliberate design choices are embedded in this prompt:
Empathic reading instruction: “read like you’re one of them” shifts the model from neutral summarizer to engaged peer. This produces more nuanced signal extraction than a generic “summarize key points” instruction.
Subtext detection: “catch the half-said frustrations, the excited tangents” explicitly instructs the model to infer emotional and strategic signal from indirect language — something LLMs can do well when given permission.
Persona decoding: “If they rant about time drains, hear ‘AI automation!’” teaches the model to translate symptom language into solution categories. This is essentially coaching the model to think like a content strategist.
Immediate productization: each nugget comes with five ready-to-use content angles. The output is not just insights — it is an actionable content brief.
Trust anchor: “Stay true — no made-up nonsense” is a grounding instruction that reduces hallucination risk on qualitative extraction tasks.
How to Use This
- Paste a transcript (coaching call, session recording, podcast interview, or sales conversation)
- Run the prompt
- Review the 3–10 nuggets for the 1–3 that resonate most
- Use the five content angles for each to plan your content calendar for the week or month
This works particularly well on conversations that felt unstructured or wide-ranging at the time — the model often surfaces coherence and patterns the participants missed in the moment.
Extension: Make It Yours
Customize the audience lens by replacing “solo knowledge entrepreneurs” with your specific client archetype. The prompt’s logic remains intact — only the empathic reading frame changes.