The System

Everything named in the book, in one place, in case you lose the thread. Map. Overlays. A cadence. What to own, what to build, how to keep it honest as everything around it changes.

The Compass

Two questions. Answer both, and you’ve got your bearing.

One: Does this actually need my judgment? Yes or no.
Two: Am I adversely impacted if AI gets it wrong? Yes or no.

The answers point to one of four directions:

EXECUTE — hand it over. AI does it, you don’t look back.

REVIEW — AI does the work; a human reads it like a stranger’s work before it goes out, because the day it’s wrong is expensive.

COLLABORATE — think with it. The cost of a bad session is a wasted hour; the upside is thinking you couldn’t have reached on your own.

STEWARD — the call is yours. AI gathers, organizes, drafts, argues, shows you what you’d miss. Then you decide, and then you answer for it.

None of the four says “don’t use AI,” and none says “do it by yourself.” You are never replaced, and you are never alone. There are no wrong answers — only the wrong ratio. And there’s a third thing no bearing can see: meaning. Some work you keep because doing it yourself is the point. That one only you can declare.

The Map

Mark a destination — a real thing you want to reach. Break the way there into legs, and list them. Take a bearing on each leg with the two questions, and label it with its direction. The list reads like the dotted footpath lines on a hiking map, each leg labeled with how you cross it.

AI walks every leg you assign it. The trip shrinks to the legs you keep for yourself. That is running at the speed of light. The checkpoints are the calls with your name on the outcome — the legs that matter too much to delegate.

Map Stacking

Keep the Map when you reach what you marked. Do that with every Map you finish and they stack up. A stack of maps is an atlas, and yours becomes a record of how you travel: which legs you keep, which you hand off, and what you decided along the way and why. Every finished Map cuts the cost of the one after it. That practice is Map Stacking.

map overlays

Five transparent capability sheets laid over your Map, on the repeatable legs you assigned away:

GENERATION — drafts, writes, designs, brainstorms.

RESEARCH — gathers, summarizes, monitors, checks. The conclusions you draw from the pile are still yours.

PRODUCTION — assembles the finished pieces.

DISTRIBUTION — publishes, posts, schedules, delivers, follows up.

DECISION-SUPPORT — watches the numbers, flags the anomalies, and frames your choices. Frames. Not makes.

Cutting through all five: the checkpoints — defined points where the work has to pass through a named human, you, before it touches the world. Capability without checkpoints isn’t an advantage; it’s a system that sounds right and is wrong somewhere you haven’t found yet.

Build your top four highest-impact decisions from your Map, and nothing else until those four are handled. What you build is sized by what it serves, not by what’s technically possible. Grown, not bought.

A cadence

Automation is never finished, because the work it runs on never stops changing.

WEEKLY, DURING CHANGE — twenty minutes. Is the output still right? Still right for us? What’s the weirdest thing it did this week, and do you understand why?

MONTHLY, IN STEADY STATE — same look, plus one harder question out loud: has the ground this was built on moved?

ALWAYS, ON TRIGGER — three tripwires override every calendar: results decline (especially explainably), conditions change, capabilities jump. Any tripwire fires, you review now.

The three failure modes

Every AI disaster is one of three patterns.

COGNITIVE OVERLOAD — judgment delegated due to overwhelm. A task lands at Execute correctly, grows edges, and no one looks again because the output still looks fine.

HAND WAVING — warnings get ignored. The review exists on paper and gets skipped in practice. The giveaway: a review meeting you keep trying to cancel.

RUBBER STAMPING — AI’s recommendations become decisions. The checkpoint still exists; the deciding stopped. The question that exposes it: when did you last say no to the system?

Not one is a technology failure. They’re ownership failures — the gap between a system that runs and a human who answers. The fix was never technical. It’s a person.

The Five Functions

Five things stay yours no matter how good the tools get.

OWNERSHIP — Responsibility is what happens after you decide. Ownership is what happens before you do.

DISCERNMENT — Information answers: what’s true? Judgment answers: what matters here? Discernment is judgment from what you know, not what you think.

COMMITMENT — A decision is what you choose. A commitment is a decision that stays chosen when the conditions it was born in change.

CARE — Care is not the performance of understanding. Care is being changed by what happens to another person.

IMAGINATION — Not the ability to generate possible futures. The taste to see which one deserves to exist, and the craft to make it real.

The Builder and the Spectator

The Builder converts intention into an attempt — something that exists outside their head and would be judged by a stranger. The Spectator converts intention into an inquiry: one more course, one more article, one more “once things settle down.” No one is born a Spectator. It’s a habit, not who you are.

The Permission Lane

Idea → Research → Design → Build → Distribute → Sell. Six stops. Having the idea was always free; every stop after it cost money, time, energy, stress, or permission. Most ideas died at stop two or three, unwitnessed. Almost every barrier on the lane just came down.

Echelons Above Reality

The altitude where decisions feel free because the people making them never face the consequences. Every organization has one.