AI-Leverage
← Writing

2026·04·22 · Essay

What leverage actually means.

The word has been flattened. People now use leverage the way they used to use synergy — as a vague intensifier attached to anything that sounds good. A business coach tells you to leverage your network. A SaaS landing page tells you to leverage AI. A productivity guru tells you to leverage your mornings. None of these sentences carry any information, because the word has been drained of the specific thing it used to mean.

The original use of the word is mechanical, and it is worth returning to. A lever is a simple machine that lets a small input force produce a larger output force, in exchange for the force being applied over a greater distance. You push on one end, the other end moves a load that your unaided hand could not shift. Leverage, in this original sense, is not more effort. It is a specific configuration of effort, where the point of application and the geometry of the system matter more than the size of the push.

In a business, leverage names the same idea. There is a place — usually one, sometimes two — where a change of a certain size produces a change several times larger than its size, somewhere downstream. The rest of the business has changes too, but those are more linear: push harder, get a little more. Leverage is the place where push is non-linear. A small improvement there reorganizes how many other things work.

That is why the word is worth guarding. “Leverage your morning routine” is almost never leverage in this strict sense. It is a productivity tactic, and the output is roughly proportional to the input. The mornings do not become a lever unless they happen to sit at a point where their output reorganizes something much bigger — and usually they do not.

So what does real leverage look like in practice? It looks like this.

A consultancy I worked with had twelve people. Their proposal process took five to seven days per proposal. The founder rewrote every draft at least once — sometimes twice — because the writing coming back from the team was close but not quite the thing the client needed to read to say yes. The founder was the bottleneck, but she also could not stop being the bottleneck, because the quality depended on her. Everyone knew this and it felt unsolvable.

When we looked at the actual workflow — which is almost always the first move — it turned out the leverage was not where anyone expected. It was not in making the team “better writers.” It was not in training them with AI tools. It was not in the founder writing faster. The leverage was in a single asset that did not yet exist: a short document that captured how the founder actually made editorial decisions on a draft. What she reached for, what she cut, why a sentence that was technically correct was still wrong for that particular client, what tone came after a specific kind of discovery conversation.

Once that document existed — and it took about four hours to write, with her — two things became possible. First, the team could draft in her idiom on the first pass, instead of on the third. Second, an AI working alongside a team member could now produce a first draft that was recognizably hers in shape, because the model had something real to be faithful to. The proposal cycle went from five to seven days to two to three. More importantly, the founder stopped being the last line of defense on every proposal. She became the first line of direction on the highest-stakes ones and left the rest alone.

Notice what leverage was not, in this case. It was not “AI.” It was not “productivity.” It was not “training the team better.” Those are all things that might have been done and none of them would have produced the change. The leverage was a specific, small, missing piece — one document of about 1,200 words — sitting at a point where its absence was forcing the founder to be involved in every cycle. Once it existed, the geometry of the whole system shifted.

The generic version of this story would have been: “We used AI to speed up their proposal process.” That sentence is technically not false. But it hides the only thing that mattered, which is that the leverage was in a structural gap, and AI was the medium that made closing the gap useful. If the gap had not been found first, using AI would have produced more proposals faster, in the same not-quite-right voice, and the founder would still have been rewriting every one.

Leverage is the place where a small structural change produces a larger change elsewhere. If you cannot name the structural change and cannot name where the larger effect appears, you are not looking at leverage yet. You are looking at activity.

This is why “leverage” matters as a discipline, and not just a word. If I hold the strict meaning, I cannot let a client pay me to produce more activity. The work has to find the specific place, or there is nothing to do. If I use the loose meaning, the work becomes indistinguishable from generic consulting, and I have no way to tell whether any of it helped.

Two more things follow from this.

The first is that leverage is almost never obvious. The person whose business it is usually cannot see it, because the thing that is in the way of them seeing it is the same thing that is in the way of them doing it. If the founder could see the missing document, she would have written it years ago. It is the kind of thing that is invisible until someone outside the system looks at how the work actually moves. That is why the first session — the Discovery — is almost entirely diagnostic, not solution-shaped. The solution becomes obvious once the diagnosis is right. Forcing a solution before the diagnosis is right is how a lot of work produces motion without movement.

The second is that once you have the discipline of looking for leverage, you stop being able to do a lot of work that other people will sell you. You stop being impressed by “we automated your intake form.” You notice that the intake form is a surface. You ask what the intake form is doing in the business — what the next five things that happen after it are, what breaks if it gets faster, what stops breaking, what no one noticed was being done by the slowness. Sometimes the right answer, surprisingly, is to leave the intake form alone and fix something else entirely. That is what the discipline feels like from the inside.

I should end with what all of this has to do with AI specifically. The short version: AI has made the search for leverage both harder and more rewarding. Harder, because AI looks like leverage almost everywhere — almost any task can be made faster with it, which is a problem, not a gift, if you are trying to tell real leverage from lookalike leverage. More rewarding, because where real leverage does exist, AI often lets you close the gap with a change that was structurally impossible five years ago. The missing document does not just sit on a server. It lives in a way that can be consulted, extended, refined as real drafts go through it. The geometry of the lever becomes richer than it could have been before.

But the word is worth guarding. If leverage gets used to mean anything useful, it stops meaning anything at all — and we lose access to one of the few words that still points at something specific.