What AI Is Changing: The Things That Break When the Technology Works.
On a Tuesday morning someone in the pricing team sends round a piece of analysis that would have taken the head of pricing most of a week. It is good. Not brilliant, but good, and it took her an afternoon.
He reads it twice. The second time he is not checking the numbers. He is working out what he is for.
That moment is the part of AI change that never makes it onto the program plan. The plan tracks licenses, training completion and usage. It has no row for what just happened to the most experienced person in the room.
AI is changing the workplace, though not in the way the headlines suggest. Jobs are not disappearing wholesale and departments are not being automated overnight. What is changing is smaller and harder to see: how decisions get made, where expertise sits, and what it means to be good at a job. Those three things are what most organizations are built on, and none of them appear in the business case.
What the technology moves.
Start with the work itself.
The technology takes over tasks that used to require judgment and surfaces answers that used to require experience. A question that would have gone to the person who had seen it before now gets a competent answer in thirty seconds. That answer is usually right. Occasionally it is wrong in a way only the person who had seen it before would catch.
So three things move at once. Decisions drift toward whoever is holding the tool. Expertise stops being the thing you go and ask for and becomes the thing you use to check what you were handed. And being good at the job stops meaning you know the answer and starts meaning you know when the answer is wrong.
That last one is difficult to hire for, difficult to train for, and almost impossible to put on a performance review. Most organizations have not tried.
The identity problem nobody puts on the program plan.
When a senior person's expertise stops being scarce, something happens that no training session addresses. They do not lose their job. They lose the thing that made them the person you asked.
That is an identity shift, not a technical one, and identity does not move because someone sat through a two-hour session on prompt writing.
Look at who picked the tools up first in your own organization. It is rarely the people with the deepest expertise. It tends to be the ones who already had high agency and little fear of looking incompetent. They experiment, they fail, they learn, and nothing about their standing depends on already knowing the answer.
The people who hang back are usually protecting something: twenty years of hard-won judgment, a reputation for being the one who knows, a settled sense of what they are worth here. Using the tool badly in front of a colleague puts all of it at risk. What they are avoiding is exposure.
The expertise inversion.
Now run that forward a few cycles.
If the people with the least experience are using AI heavily and the people with the most are holding back, the distance in capability does not close the way you would expect. It inverts.
The junior analyst produces more, faster, across a wider range of problems. The senior one produces at roughly the rate they did last year. Work starts getting routed to whoever moves fastest, and the person with the deepest judgment becomes a bottleneck to be worked around rather than a resource to be used.
You can see it early if you look for it. The senior person stops being copied into the threads where the work gets decided. Their calendar empties of the interesting problems and fills up with approvals. They are still the most capable person in the function, and they are being used as a rubber stamp.
Nobody decides this. It emerges from a thousand small routing choices made by people trying to hit deadlines, and by the time it is visible on an org chart it has already changed who talks to whom.
The missing rung.
Here is the part that shows up years later.
Junior people used to become senior people by doing work that was, frankly, a bit dull. Reconciling the numbers. Drafting the first version. Sitting in on the call and writing it up afterward. Nobody enjoyed those tasks and everybody learned from them. That was the apprenticeship, even in organizations that never used the word.
Those are precisely the tasks the technology is best at.
So the rung people used to climb from is going. You can still hire juniors and you can still hire seniors, but the path between the two is thinning, and the organizations pulling the most productivity out of AI today are often the ones removing it fastest. In ten years they will be trying to hire a level of judgment that nobody was given the chance to build.
It is already visible in hiring. A team that would once have taken on three juniors takes one, because the work those roles existed to absorb now happens in the tool. On this year's numbers that reads as efficiency. It is also a decision about who will be capable of running the function in a decade, made by someone optimizing a quarterly budget.
This is what we call the missing rung, and it is one of the central arguments in The Elephant in the Algorithm. None of it is an argument against using the technology. It is an argument for being deliberate about what the technology takes away from people on their way up.
What this asks of a leader.
None of this is solved by buying differently or training harder.
It is solved by being explicit about things most organizations leave implicit: who is learning what and from whom, which tasks exist because they produce output and which exist because they produce people, and what a senior person is for once their recall advantage has gone.
That last one matters more than it sounds. The senior people reading the room right now are making private decisions about their own future, and they are making them without any information from you.
The principle we work to is raise the human rather than remove the human. You can deploy AI so that it takes work off people, or so that it raises what people are capable of. The tools do not care which. The design choice is yours, and in most organizations it is currently being made by default.
If you want to know how this is landing in your own organization, the AI Alignment Snapshot gives you a read on it in about eight minutes. It is free, and it asks nothing of you except honest answers.
And if you want to talk through what you are seeing, book a discovery call here.
Take it with you
Download this as a PDF
A clean, branded version to read offline or share with your team.