The Apprentice Was Asked to Arrive Finished
To the class of 2027:
Congratulations. You have been invited to begin at the bottom, assuming you can demonstrate that you already know how to climb.
This is not a joke. It only has the shape of one.
UBS will ask its graduates and interns to prove they can use AI to improve outcomes and efficiency before it lets them into Global Banking and Markets. I picture the interview room: the clean water in the little bottle, the bright young face, the panel across the table asking the question in its well-ironed language.
Tell us how you used the machine.
Tell us what it made better.
There will be a right answer, naturally. Not I asked it to write a cover letter because I was scared and rent was due. Not I used it to turn three hours of spreadsheet work into forty minutes so I could look out the window and wonder whether this was a career or a hostage exchange.
No. They will want the grown-up answer. A concrete task. A measured gain. A human being who can explain where the tool helped and where the human hand stayed on the wheel.
Fine. That is not an insane thing to ask. A bank should not hire people who treat a language model like a Ouija board with a better user interface. If you are going to hand a machine confidential material, numbers, client notes, the soft underbelly of somebody else’s money, you ought to know what it is doing and what it is not doing.
The trouble is not the question.
The trouble is the timing.
An apprenticeship used to mean the world admitted you were not ready yet. That was the bargain. You arrived green, carrying a degree and a shirt you hoped looked expensive enough. Then you did the rotten small work. You checked the numbers. You built the deck. You made the bad first version. You got corrected by someone older, perhaps cruelly, perhaps usefully, sometimes both. You learned which neat answer had a dead rat inside it.
Nobody enjoys this part. I certainly did not enjoy learning anything. The post office did not hand me a philosophical pamphlet and a wise mentor. It handed me trays, deadlines, men who had mistaken authority for a personality, and enough wrong addresses to teach me that a tiny error can ruin somebody else’s afternoon.
But the work was teaching me while it was wasting my time.
That is the miserable beauty of being a beginner. You have permission to be slow in the only place slowness can turn into judgment.
Now the banks are feeding that beginner work to machines. Research. Financial analysis. The presentation that gets made at midnight because somebody senior wants to move four arrows two millimeters to the left. The things that were boring, yes, but also the things from which a young analyst learned that the client number in cell F19 did not smell right.
Then they look at the empty space where the apprenticeship used to be and say: show us you can use the thing that emptied it.
It is a neat little circle. You cannot enter without fluency. You cannot gain fluency without getting near the work. And the work has been placed behind glass with a bot in front of it, smiling like a concierge.
UBS has an internal AI Fluency Pathway for the people it hires. The name has the polished, toothpaste quality of modern corporate language, but I will give them this: it is better than pretending the problem does not exist. Somebody in the building has noticed that a person who lets software do the first ten years of thinking may not have much left when the software is wrong.
They even plan to teach responsible use.
Good. Teach it hard.
Teach the new kid that a sentence produced in half a second can be false in six different ways. Teach her that a model can make a beautiful chart from bad data without breaking a sweat, because it does not sweat and the chart is not its mortgage. Teach him that efficiency is not the same thing as understanding, and that confidence is not the same thing as a reason.
But a pathway is not a pathway if it begins after the bridge has been removed.
The old junior work had a hidden function. It gave you cheap mistakes. You could be wrong in a room where somebody caught it. You could ask the stupid question before the stupid question had a client’s name on it. You could spend four hours rebuilding a model and discover that the two columns were using different dates, then remember that pain forever.
The machine removes the pain. Very generous of it.
It may also remove the memory.
There is a word going around for this: never-skilling. A nasty word. It sounds like a disease caught from a keyboard. The idea is simple enough. If the tool does the hard early parts before a worker has a chance to struggle with them, the worker may become fast without ever becoming capable. They can operate the dashboard. They can speak the new dialect. They can produce the answer. But when the answer is a lie, they have no old bruises to tell them where to look.
That is not a problem unique to bankers. It is coming for every job that once had an entry level. The lawyer’s clerk. The junior coder. The assistant editor. The young accountant. The kid in the shop who used to learn by making an imperfect thing beside someone who knew how to fix it.
The companies will say the work is changing. They are right. Work always changes. I am not writing this with a quill while a horse waits downstairs. I use machines. You use machines. The question is not whether the tool belongs in the room.
The question is whether anybody in the room will still know how to work when it goes quiet.
So, class of 2027, learn the AI tricks. You have to. The gatekeepers have put the trick in the lock.
But do not confuse the key with the place it opens.
Ask to see the ugly spreadsheet. Ask why the number is there. Do the long version once, even when the machine offers the short one. Make a mistake small enough to survive. Find an older person who will tell you when you are bluffing, and buy them coffee if necessary. Those people are disappearing faster than the junior tasks.
One day a client will ask a question the box has not rehearsed. The room will turn toward you because your name is on the work.
By then, fluency will not save you.
You will need to know what the words mean.