Dr Jim Kennedy  ·  Build Useful Leverage

Make It Think Like You

The operator's starter for building your expertise into AI — works in any model, scales all the way to Claude Code.

You came here to find out which AI to pay for. Here's the honest answer — and the more useful one.

Which model barely matters now. Tested head-to-head, the big four have converged; on most real work the differences are subtle. Everyone has access to the same models. So if the model were the edge, everyone's work would be world-class — and it plainly isn't. The edge isn't the tool. It's what you put into it.

Anyone can type a prompt. A prompt tells the AI what to do once. Encoding tells it who it's working for — every time, without being asked. That single move turns a generic chatbot into something that thinks like a 20-year operator: you.

It works in whatever you already use — ChatGPT, Claude, Gemini — and it's the foundation of the most powerful way to run AI: Claude Code. This page isn't a description of that. It's the build. Three prompts, five inputs, one habit. You can have the first version running before your coffee's cold.

Domain depth × AI capability = leverage.

Technical skill is now a commodity. Your judgment, encoded, is the moat. The model is the engine — your context is the fuel. Most people are driving a Ferrari on an empty tank.

The five things the AI has to know to think like you

These five are the whole game. Brief them like you'd brief a sharp new hire on day one — plain language, no jargon.

  1. 1
    Who you are, and what you actually do. Your role, your field, how long, and who you serve.e.g. "Civil engineer, 22 years. I design and de-risk mid-rise structures for commercial developers."
  2. 2
    How you make a call. What you weigh first, second, third — and the question you always ask before deciding.e.g. "I check the downside before the upside. I never trade safety margin for speed."
  3. 3
    What you'd never sign off on. Your red lines. The standards you won't cross. The mistakes you've learned to catch. This is your 20 years talking — the part no competitor and no model can copy.
  4. 4
    How you sound. Direct or warm, plain or technical, short or thorough — so the output sounds like you, not a robot wearing your name badge.
  5. 5
    What "good" looks like. One example of work at your standard — paste it, or describe the bar. That's the target the AI aims at.

You don't have to write a word of it

Here's the upgrade. Most people read a list like that, mean to fill it in "later," and never do. So don't fill it in. Make the AI do it.

Open whatever AI you use and paste this:

Prompt 1 — Build my profile
Interview me so you can work like a 20-year version of me. Ask me one question at a time about: (1) who I am and what I do, (2) how I make decisions, (3) what I'd never sign off on, (4) how I want you to sound, and (5) what 'good' work looks like to me. When you have enough, write it back to me as a clean profile I can save and reuse. Start with question one.

Answer like you're talking, not writing. Ten minutes of back-and-forth and it hands you a finished profile — sharper than the one you'd have stared at a blank page trying to write. Read it, fix anything that's off, and you've got the asset. That's it.

Now put it where the AI reads it every time

This is the step that separates you from everyone retyping context into a fresh chat all day. Don't paste your profile each session — store it once, in the right place.

A rule worth keeping: what's loaded every time stays short; everything else loads only when it's relevant.Pack everything into one giant always-on instruction and you choke the AI on noise. Keep a short "how we work together" core, and let the meaty stuff sit in files it pulls in on demand.

Then make the file sharpen itself

Most people retype the same correction forever. You're going to write it down once — and you're going to make the AI do that too.

End any real working session with this:

Prompt 2 — Compound it
Based on how this session went, what did you learn about working with me that we should add so it goes smoother next time? Suggest the exact wording, and tell me whether it belongs in my always-on preferences or in a separate context file.

Every wrong turn becomes a rule. Every rule makes next week's output better. Within a month the AI feels like it was built for your business — because it was. That growing file, not the model, becomes the thing nobody can take from you.

And when you catch yourself doing the same multi-step job a second time, capture it:

Prompt 3 — Turn repeat work into a reusable move
We just did [the task] well. Look back at how we got there — the steps, the questions you asked me, what 'done' looked like — and write it up as a repeatable procedure I can trigger next time instead of explaining it again.

That's the difference between renting intelligence and owning how it works.

Where this goes

You've just built the first layer. There are four, and each one only works once the one before it is solid:

  1. 1Context — it knows you and your business.(You just did this.)
  2. 2Connections — it can reach your real world: calendar, tasks, email, files, the numbers you check every week.
  3. 3Capabilities — it knows how you do the work: your repeatable procedures, in your style, to your standard.
  4. 4Cadence — the right things start happening on a schedule, without you asking.

Do the first layer and whichever AI you use stops feeling generic. Climb all four and it can run real work end to end — which is exactly how I run mine: on Claude Code, as a non-coder.

One honest caveat for senior operators:give it reach in stages, not all at once. Treat it like teaching someone to ride a bike — hand on the seat first, let go only once it's earned trust on smaller things. And remember that an instruction not to do something is weaker than simply not handing over the key: be deliberate about what you let it touch. You can outsource the work. You can't outsource the understanding — your judgment stays in the seat.

This is a starting point

The aim is larger than making an AI sound like you. It is to make the judgment you have earned visible, reusable, and capable of creating something worthwhile.

The starter above works on its own. Keep improving it every time real work exposes something the system does not yet understand.

— Dr Jim Kennedy

Get The Signal

One useful idea each week about what AI is changing — and how to make your experience count for more.

No spam. One email a week.