Three AI Prompts to Extract Expert Knowledge From Your Team

Three Prompts to Capture What Only One Person Knows

TLDR: This article provides a three-phase framework using AI to extract expertise from key employees, identify automation opportunities, and create reusable prompt templates. The approach requires no technical setup - just copy-paste prompts into ChatGPT, Claude, or Gemini.

Every organization has that one person who does something brilliantly. They quote timelines instantly. They handle difficult clients effortlessly. They produce estimates in twenty minutes that take everyone else three hours. The problem? All that knowledge lives nowhere except inside their skull.

This is knowledge concentration - the silent growth killer in most companies. When that expert goes on holiday, work slows. When they get promoted, their replacement struggles for months. When they leave entirely, years of accumulated wisdom walk out the door with them.

The article identifies two related symptoms. First is tribal knowledge: expertise one person holds that nobody else can replicate. The example given is compelling - a commercial roofing company with 25 estimators where one person with 22 years of experience produced accurate estimates in 20 minutes while everyone else took three hours and hit 75% accuracy. Second is what the author calls the "AI gap": some employees use AI to move three times faster while the rest watch from the sidelines, unsure where to start.

The proposed solution is refreshingly practical: use AI to interview your experts, extract what they know, figure out which tools can automate parts of the process, and turn everything into prompts anyone can use. Three phases, no coding, no consultants.

The first phase focuses on extraction - getting everything out of your expert's head into a document anyone can follow. The key insight is giving AI a clear stopping point. Without constraints, interviews wander. With a 20-question limit, AI stays focused and covers what matters. You copy a prompt into your preferred AI tool and answer questions one at a time. The AI does the structuring.

For architects and team leads, this approach addresses a real organizational risk. Single points of failure in knowledge aren't just inconvenient - they're architectural debt in your organization. The same way you'd refactor a codebase with too many responsibilities in one module, you need to distribute critical knowledge across your team.

The three-phase output is practical: a documented Standard Operating Procedure extracted from your expert, a clear view of what can be automated and which tools to test, and a shareable prompt template anyone on the team can use. You can run all three phases or stop after phase one if that's all you need.

What's missing from this approach? The article doesn't address how to validate that the extracted knowledge is actually correct and complete. Experts often have unconscious competence - they do things without being able to articulate why. An AI-structured interview might capture the what without capturing the deeper why. Teams should plan for a validation step where the documented procedure is tested against real scenarios.

Key takeaways:

Knowledge concentration is a growth killer - critical capability shouldn't sit in one head
AI can conduct structured interviews to extract expertise without wandering
The 20-question constraint keeps extraction focused and actionable
Output includes documented SOPs, automation opportunities, and reusable prompt templates

Tradeoffs:

Structured extraction gains efficiency but may miss tacit knowledge that experts can't articulate
AI-generated documentation increases accessibility but requires human validation for accuracy
Prompt templates enable knowledge sharing but may oversimplify nuanced decision-making

Link: Three Prompts to Capture What Only One Person Knows

This article summarizes content from the AI Adopters newsletter. The information presented here is based on newsletter content and may not reflect the complete original articles.