Michael Gardon · CHIEF AI Accelerator
Accelerator / Curriculum / 0.1 Mental Model
Phase 0 · Foundations

Congrats — you just hired a brilliant new Chief of Staff

Get the working model that makes AI compound for you.

Foundations
The Mental Model

Your new hire

You just hired a Harvard MBA, the sharpest researcher, writer, and analyst you've ever worked with. It can run circles around anyone in your office when it comes to execution: a true Excel ninja, the best PowerPoint designer, and it knows every tool or system you could possibly throw at it. AND it can code, like really code. Not just code, but design and execute systems at speed.

There's just one catch: it's also basically an insecure millennial. It walked in this morning knowing nothing about you. Not what you want, not what matters to you, not where anything lives, not your standards. It needs constant feedback and validation.

But the good news is that it's an absolute beast of a learning machine, one that eventually can make your life sane and your performance world class.

The question is: how would you leverage and cultivate a promising new hire?

That's the whole mental model. Every framework in the program, prompting, projects, context, delegation, is a different answer to one question: how do you make a brilliant-but-new Chief of Staff useful?

The Goal

Compound your time value

Why hire a Chief of Staff? To make your life easier, so you're freed up to do what you do best, and do what others can't do. Otherwise, what's the point?

Same with AI. Execs who invest in learning AI now, through building, not only become more strategic, unlock their own creativity, and execute faster. They also understand how to push AI usage down through their organizations. That payoff helps the whole org move faster and more efficiently.

The Frame

The relationship progression

Think about how you'd onboard a Chief of Staff.

Setup

Getting their tools and credentials, showing them where to find the right information, describing workflows.

Trial

You give them some basic tasks like research, right out of the gate. Scoped work you're pretty sure they can handle. Nothing crazy.

Closer collaboration

You bring them into the fold. Daily meetings where you download context: what's going on, what's important this quarter, what you need today. You start bringing them in on key deliverables. Scope is still tight, and they're getting better at understanding how you do things and what you like, your taste and judgment.

Delegate

You let them run and execute autonomously in their wheelhouse. They start taking initiative. Feedback loops get tighter, their capabilities expand, your life gets easier.

This is exactly how the AI relationship matures. Except, to work, you need to encode what you want, how you want it, and how you'd do it, in files.

The Model

The executive AI progression

R1 R2 R3 Skills compound · relationship matures →
01 · Tool

Scoped single tasks. One prompt, one answer.

"What is the answer to this question?"

You describe what you want; AI produces information. Most people start and stop here.

Skill installed: prompting · how to scope · the 4 C's
02 · Collaborator

Iterative, multi-step work, judgment shared.

"Help me think through how that answer affects my business."

Research, analysis, drafting. You encode taste, voice, and process so the AI references them automatically.

+ Context architecture · Projects · Reusable skills
03 · Delegate

Agents act on your behalf, with verification.

"Build a system to deploy our solution and monitor its effects."

Bounded autonomy inside rules you set. You keep the boundary; the system runs. Only works because Ring 2 came first.

+ Connections · Scheduled tasks · Agent design

Each ring is a transfer of a different kind of judgment. Ring 1, you transfer the task and keep all judgment. Ring 2, you transfer process and taste, and keep the decision. Ring 3, you transfer decision authority within bounds, and keep verification.

The Executive Shift

Four habits that come straight from the frame.

Most executives treat AI like a search box, then complain when the output is generic. The Chief of Staff frame fixes that without teaching you a single prompt trick. Treat AI as an employee: ask, spar, iterate, encode.

Encode context as the asset

Every minute you spend giving context, the AI repays in output quality. Move from prompted context to encoded file context. This is the highest-leverage piece. You are encoding your taste, preference, judgment, and what matters to YOU.

Be radically specific

"How do I automate my marketing?"

"Think like a system architect. I need a high-level brief on how to automate my email marketing campaigns. Include the top-level system and sub-components. Flag any needs from a human as blockers, and provide time and cost estimates so I can decide how to proceed."

Make it ask you for what it needs

Tell it to ask you questions about scope. Then tell it to capture the context for later encoding.

Ask it how to do things outside your comfort zone

Non-technical? Ask for step-by-step instructions and to define terms you don't know.

(I have no technical experience, and every single day I'm using Terminal and building code, just because I ask. I also fixed my lawnmower and took apart a carburetor.)

Expectations Reset

AI doesn't need to level up. You do.

It's hard to comprehend what AI is capable of right now. Most execs want to jump straight to automation, but this is a big trap, because your capability set isn't there. And by capability, I don't mean your technical ability. That doesn't really matter anymore. I mean your ability to extract your knowledge and judgment into English for the benefit of a system. Turns out this is hard for many execs.

I know that's a lot to take in. But learning to operate with a machine, to scope, chunk, guide, direct, evaluate, and decide, can completely transform the way you work and your results for the better. You will get there. You may not be slinging code or building a fully autonomous organization, but you'll be able to delegate exactly what you want to someone who can execute automation with AI.

The target isn't to push everything to Ring 3. It's to be able to operate in the native AI future. This wave is moving faster than any technological advancement in history, and you need to know how to operate in it. This program will get you there.

Next in the Sequence

You have the model. Next, learn to brief.

Before the prompting craft of Ring 1, one short reading bridges the philosophy and the tactics: Dr. Robert Winter on using AI to surface assumptions without outsourcing your judgment.

Open the Winter Method →

Optional reading

Curious how it actually works under the hood? How AI thinks & remembers →