Operating models fail because we only document processes (Explicit Knowledge), not judgment (Tacit Knowledge).
In this exercise, you'll extract your own "gut instinct"—a decision you make intuitively—and transform it into a System Prompt that could run an autonomous agent.
Philosopher Michael Polanyi famously said: "We can know more than we can tell."
The shift: AI is a Tacit-to-Explicit converter. You can now "clone" expertise by capturing the reasoning patterns that never make it into a handbook.
Pick a repetitive but complex decision you make often. Examples:
Be messy. Explain why you reject things, what "smells wrong," and what "looks good." Don't try to be structured—just capture your intuition.
You're moving a piece of your brain (Intrinsic Load) into code (Extraneous Load), freeing you up for higher-level work.
The validation test: Does the System Prompt Gemini generates look like a rulebook for you? If yes, you've just digitised your intuition.
Using Sweller's Cognitive Load Theory, identify what parts of your team's work should be handed to AI.
The goal: Eliminate Extraneous Load so teams can focus 100% on Germane Load—moving from "Code Monkeys" to "System Architects."
The audit will categorise your tasks into three buckets:
The insight: How much time is your team wasting on "noise" that AI could eliminate?
Psychological reframing for teams facing AI-driven change. Click each card to flip it.
"I spent 10 years mastering manual masking and colour grading. Now a one-click tool does it. My skills are obsolete."
[CLICK TO FLIP]
"Your hands are no longer required, but your eye is essential. The tool can generate, but it cannot judge.
We need you to shift from 'Operator' to 'Director'—using your taste to guide the AI, just as a Creative Director guides a junior team."
[CLICK TO FLIP BACK]
"I can't keep up. Every week there's a new model—Llama, Claude, Gemini. I'm drowning in API documentation."
[CLICK TO FLIP]
"Stop trying to learn every tool. That is the Platform Team's job.
Your job is to learn the patterns (RAG, Agents), not the pixels. We abstract the complexity so you can focus on the business logic."
[CLICK TO FLIP BACK]
"If we implement this, my team shrinks from 20 people to 5. I will lose my status and budget."
[CLICK TO FLIP]
"In the AI era, we measure leaders by 'Output Velocity,' not 'Headcount.'
You aren't managing a smaller team; you are managing a significantly higher-throughput system. You are now running a factory, not a workshop."
[CLICK TO FLIP BACK]
Based on Skelton & Pais's framework, adapted for scaled intelligence:
Don't wed the company to one vendor. The best architecture builds a highly scalable middle layer that:
The goal: Model Agnosticism. Your operating model should survive vendor changes.