The shift
The price of trying has collapsed.
For two hundred years the size of an ambition was set by how many people you could put behind it. That is no longer true, and it changes what is worth measuring.
One person, the work of hundreds
01What changed
The distance from an idea to a working version of it is now a conversation.
Every organization keeps a list of things it decided not to do. Most of that list was never written down.
The reason was almost always the same. Finding out whether an idea was any good cost more than the answer was worth. So it was declined in a meeting, and the meeting was right.
That price has fallen. One person can describe what they want in ordinary language and get back something that runs. The same person can direct machine workers that read, draft, analyze, build and test at once, for the cost of electricity.
Since the first command line, every tool charged the same tax: you translated what you wanted into what the machine would accept. For a large class of work, that tax is gone. Remove a tax that stood for fifty years, and the activity it was holding back turns out to be far larger than anyone guessed.
So this is a fact about ambition rather than technology. If one person can attempt what used to take hundreds, the limit on your organization moves. It stops being capacity and becomes nerve, judgment and taste. Every competitor rents the same machines, at the same posted prices, on the same morning. Which problems you choose is the difference.
Eight declined · one now open
What to do this week. Ask three of your leaders what their area declined in the last three years, and what trying would have cost then. Who: you, by email. How long: half an hour to read the replies. You end up with three declined projects and their old prices, and it is done when each one carries a name and a number.
02Where the work goes
The hard part does not disappear. It moves up a step.
Every real abstraction takes difficulty away at one level and puts it back at a higher one. The difficulty climbs, and so do the people who were carrying it.
Compilers removed the need to write in the language of the machine, and that understanding was lost, exactly as the critics said. The programmers who were freed then built operating systems and networks their predecessors could not have imagined. The lost depth was real. The gained breadth was larger.
The freed hours were not idle hours. They went into the next level of difficulty. An amplified organization is not an easier place to work. Its people spend their days on the questions the mechanical work used to crowd out. What should exist? For whom? At what standard? And who answers for it?
Those questions are harder than the mechanical work was, and they are the only work that stays scarce. You cannot see the value of a higher step from the step you are on.
The work rises with the tools
What to do this week. Take one job your team does weekly and draw a line down a page. Put the mechanical work on one side and the judgment on the other. Who: whoever does that job, with you in the room. How long: forty minutes. You end up with one page showing which side the hours are on. It is done when the person who does the work agrees the page is right.
03What the evidence shows
Almost everyone has the tools. Almost nobody has made the change underneath them yet.
Five published measurements. One shows what the work can now do. The other four show how much of it is still there to be spent.
64%
of American businesses using AI report no institutional change at all. Training people and redesigning how the work is done each reach about 15 percent.
U.S. Census Bureau · November 2025 to January 2026 Evidence
7%
of firms in the euro area call their use of AI significant, out of about 70 percent reporting some use. The ceiling is the same in every size class.
European Central Bank · about 6,000 firms in twelve countries Evidence
4 of 5
root causes of failed AI projects are about the organization and its people, not about the technology.
RAND · 65 interviews with practitioners Evidence
25%
faster, with 12.2 percent more tasks finished and higher quality, on work that sat inside what the model can do.
Pre-registered field experiment · 758 consultants Evidence
3%
time saved after two years, with no measurable change in earnings or hours. The same records rule out any effect on pay or hours larger than 2 percent.
Linked employer-employee records · about 25,000 Danish workers across 7,000 workplaces Evidence
Read the last two cards together. On single tasks inside what the model could do, the work ran 25 percent faster, with more of them finished and the quality higher. Across whole organizations, 3 percent. Nearly every organization is holding capability it has not spent, and the distance between those two numbers is how much of it there is.
The same experiment marks where the gain stops. On one task chosen to sit outside the model’s range, the same people were 19 percentage points less likely to be right. So the useful question is not whether to be careful. It is where that line runs in your own work, because everything inside it is available to you now.
Measure that line rather than ask about it. In one randomized trial, sixteen experienced developers worked on real tasks, timed rather than asked. With these tools available they took 19 percent longer, and believed afterwards they had been 20 percent faster. The gap between the clock and the feeling was 39 points. It is one small study on one kind of work, and it is set out in full with its design and its dates.
What to do this week. Put one question on the agenda of your next leadership meeting. What did we attempt this quarter that we would have declined two years ago? Who: you, in a meeting you already hold. How long: fifteen minutes of the agenda. You end up with a written list, and it is done when every item on it has a name against it.
04The question to carry
One question, and the rest of this site is built to answer it.
What could your organization attempt now that it could not attempt eighteen months ago, and what would it take to attempt ten times more?
Any reading of an organization makes a bet about what will be scarce. Ours is that competence is becoming the abundant part: available at professional standard, on a monthly subscription, to everyone you compete with.
The scarce things are further up the stairs. Knowing which problem is worth solving. Recognizing the good answer among forty plausible ones. Putting your name on a decision you cannot fully justify in advance. Those things were always what mattered, buried under the labor of execution that filled most of a working life.
The machines are removing that labor. What is left is the part that was always the point. So read your organization by what it has already done. Who carried the last thing you attempted, how many people would that work have needed before, and what did it cost? Each answer is a fact you can check.
Where this goes next
The six areas. What decides how much your organization can carry, and what each area has to have in it before you can attempt more.
The reading. How the reading is built, area by area, with one sample organization worked through line by line.
What to do this week. Write the question above at the top of a page and answer it with three real examples. If you cannot find three, that is the answer. Who: you, before anyone else sees it. How long: twenty minutes. You end up with a dated page, and it is done when each example names who carried it and what it cost.