Power When the Machine Does the Work

The reality

A founder built a firm on being the one who knew. The pricing logic lived in their head. The client relationships routed through their phone. The vendor terms, the margin on each job, the reason the Tuesday process worked the way it did, all of it sat with them and nowhere else. That knowledge was their power. When they were in the room, decisions moved. When they left, decisions waited.

Then the team put three years of quotes, contracts, and project notes into a system an AI could read. Now a manager can ask the tool what the margin was on the last eleven fit-out jobs and get an answer in four seconds, an answer the founder used to be the only source of. The founder felt it as a small loss they could not name. Their power was thinning, and the thing thinning it was the system they had asked for.

That feeling is the chapter. The kind of power that came from being the only one who knew is the exact kind AI erodes first. The question is what power replaces it, and whether the founder builds the new kind on purpose or watches the old kind drain.

Read this if

• Your authority in the business still rests on knowing things other people do not
• You have felt a quiet resistance to writing down what is in your head, and not looked at why
• You are putting your operation into tools an AI can read and wondering what that leaves you
• You watch other founders hold a room and cannot say what they have that you do not
• The word power makes you slightly uncomfortable, which usually means it is worth reading

Where founders learn power, and the problem with it

Most founders who go looking for a serious book on power find Robert Greene. The 48 Laws of Power has sold millions of copies and sits on a lot of founder shelves. It is worth reading, and it is worth reading with your eyes open, because Greene is honest about what it is. His preface does not pretend the book is nice. He writes that people who claim to opt out of power games "are often among the most adept players," that they "display their weakness and lack of power as a kind of moral virtue." He quotes Machiavelli approvingly: "Any man who tries to be good all the time is bound to come to ruin among the great number who are not good." His summary of the whole method is Napoleon's line, "place your iron hand inside a velvet glove."

Read it. Then notice what happens to those laws when the machine can read.

Take Law 11, Learn to Keep People Dependent on You. Greene means it straight. Make yourself the single point everyone needs, so no one can move against you. For centuries that was real power. For the founder above, it was their whole model. Here is the manipulative version stated plainly: hoard what you know, keep the team a little in the dark, stay indispensable. And here is why it is now a trap. The moment your operation is legible to a system, indispensability built on secrecy is the first thing to go. You cannot be the only one who knows the margin when the margin is in a database the tool reads. Worse, a founder who quietly protects their own indispensability will resist building the very systems that let the business run without them, and a business that cannot run without the founder cannot be sold, cannot be left, and cannot rest.

The durable version of the same law is its inverse. Greene has a second law that survives contact with the AI era, Law 5, So Much Depends on Reputation, Guard It with Your Life. Reputation is power that does not depend on anyone being in the dark. A founder whose team, clients, and market trust their judgment holds something no database can hand to a manager in four seconds. AI can surface what the margin was. It cannot tell the client whether to trust the person quoting the next job. That trust is built, slowly, and it is the power that scales instead of the power that traps.

This is the move for the whole chapter. Name the law, state what it really says, and put the version that lasts beside it.

The framework: three kinds of power the machine cannot read

Three unequal kinds of power, in the order they matter.

1. Reputation, not indispensability

Covered above and first because it is the reversal the rest depends on. The founder who stops guarding what they alone know, and starts guarding what people think of their judgment, trades a power AI destroys for a power AI cannot touch. The concrete move this week: pick one thing only you know, the pricing logic is the usual one, and write it down well enough that the team and the tool can both use it. The fear you feel doing that is the old model of power leaving. Let it.

2. Tempo, the power over the clock

There is an old observation from a British naval historian, C. Northcote Parkinson, who wrote a short essay in The Economist in 1955 with a line that outlived everything else he did: "Work expands so as to fill the time available for its completion." Give a task a month and it takes a month. Give the same task a week and it often takes a week. The founder who controls the clock controls the work. Greene makes the same point as a law of power, Law 35, Master the Art of Timing. Same lever, told from the strategist's side.

The most public example of using tempo as a lever is also the most honest about its cost. In 2017 Elon Musk told the world Tesla would build 5,000 Model 3s a week by December. It was a deliberately compressed public deadline, Parkinson's Law used as a weapon against his own team's tendency to let the work sprawl. He missed it. He missed it badly, publicly, through what he himself called "manufacturing hell," and he did not hit 5,000 a week until the last week of June 2018, roughly six months late (verified 2026-07-03, NPR/CNBC). That is the part the motivational version of the story cuts out, and it is the part that matters. The compressed deadline still pulled the work forward faster than a comfortable one would have. It also cost him public failure in front of everyone. Tempo is a real lever and a costly one. A founder who sets an aggressive internal deadline is using a genuine source of power and taking on a genuine risk of missing in front of their team. The skill is knowing which deadlines are worth that.

AI changes what tempo costs. The work that used to eat the week between deciding and shipping, the first draft, the first model, the comparison document, now takes an afternoon. The founder who understands that can compress deadlines that used to be impossible. The founder who does not will keep quoting last year's timelines for this year's tools.

3. Formlessness, the power to change shape

Greene's final law, Law 48, is Assume Formlessness. It is the one law in the book that reads better in 2026 than it did when he wrote it. His point: rigid things break, fluid things survive, and the powerful keep themselves able to change shape instead of committing to one fixed form a rival can attack. For a founder in the AI era this stops being a metaphor. The tools change every few months. The business that built one rigid process around one 2024 tool is the business that cannot move when the 2026 tool makes that process obsolete. The founder who keeps the operation legible, documented, and loosely held can reshape it as the ground moves. Formlessness here is a documented system held loosely enough to rebuild when the tools change.

A founder you might recognise

A property management firm, around 25 people, the founder eight years in and known in their patch as the person who could quote a building's running cost off the top of their head. That reputation won work. It also meant every renewal, every tricky owner, every margin call came back to them, because they had made sure it did. When the team finally built a system that held the numbers, the founder's first feeling was not relief. It was exposure. If the tool knew the running costs, what was the founder for?

What the founder was for turned out to be the part the tool could not do. Owners did not renew because a database was accurate. They renewed because they trusted the person who stood behind the number. Six months after handing the numbers to the system, the founder was spending less time being the source of answers and more time being the reason clients stayed, which was worth far more and had always been the real power. The indispensability was never the asset. It only felt like it because losing it felt like loss.

Working through it

Step 1 (this week). Name the one thing you protect. Every founder has a piece of knowledge or a relationship they quietly keep as theirs alone. Write down what yours is. Be honest about whether you keep it central because it must be, or because being the only one who holds it feels like safety.

Step 2 (this week). Write it down so the team and the tool can both use it. Whatever you named in step one, document it once, properly. Notice the resistance. The resistance is the old power leaving, and naming it is how you let it go without pretending it did not cost something.

Step 3 (this month). Pick one deadline worth compressing. Choose one piece of work where a shorter clock would pull it forward, set the aggressive date out loud to the team, and accept the risk of missing in front of them. Tempo is only a power if you are willing to spend it.

Common mistakes

1. Confusing being needed with being valued. The founder who is needed for answers is a bottleneck. The founder who is valued for judgment is a leader. AI is about to make the difference between them brutally clear.
2. Reading Greene as a manual instead of a map. The laws describe how power works, including its ugly forms. Using them as instructions to manipulate your own team builds a firm that a legible system and an honest hire will eventually expose.
3. Treating documentation as a threat. A founder who resists writing down what they know is protecting a power the machine is going to take anyway. Better to trade it deliberately for the power that lasts.
4. Setting aggressive deadlines you are not willing to miss in public. Tempo works because the clock is real. A deadline everyone knows you will quietly move is just theatre.

When to move on

Move on when your authority in the business rests on judgment people trust, when the operation is documented well enough that a new hire or a tool could run it and that makes you glad, and when you can set a hard deadline and mean it. At that point power has moved from something you guard to something you hold in the open, which is the only kind that survives a machine that can read.