How much of the work you were doing six months ago now runs without you?
Be honest with yourself about the question. I am not asking whether you use AI. Pasting text into a chatbot and getting a quick answer does not count. You were doing that six months ago. I am asking how much of your work is now automated. How many agent loops run for you every day, producing output while you do something else?
If you have a real answer, and your working day looks different than it did in January, you can stop reading. You do not need this article.
If your day looks the same as it did six months ago, this article is for you. I am going to be direct, because I think the polite version of this conversation is part of what has kept you comfortable. Whatever you have not done in the last six months, you need to start doing today.
The wind
I grew up in the UK and I walked up a lot of hills. On a bad day the wind fights you the whole way, and you learn a few things fast.
You cannot beat the wind. Pushing straight into it wastes energy you need for the climb. Standing still does not work either. You have somewhere to be, and planting your feet means you arrive nowhere, tired.
So you replan the route. You look for the path where the wind is behind you, even if that path is longer. When there is no such path, you lean in and keep walking, because stopping gets you nothing.
AI right now is that wind. It is strong, it is accelerating, and it is pushing everything in one direction. Whether you like it makes no difference to the wind. The only decision you have is the route.
The engine
Imagine you own a factory in 1900. You spent a fortune on a steam engine and on the belts and shafts that run off it. Then electric motors arrive.
You can see what they mean. New competitors start cheaper, so they keep more of their capital. The technology is new, so it will keep improving, and other people will find those improvements before you do. Ripping out the steam plant will cost you money and it will hurt. Keeping it will cost you the business.
Nobody remembers the factory owners who campaigned against electricity. The ones who survived paid the price of switching and moved on.
In your work, the engine is you. Every process in your company was designed with a person at the centre, doing the cognitive work: reading, drafting, analysing, summarising, checking.
Run the numbers on that engine. A professional on a normal salary costs an employer around 60 dollars an hour. A typical cognitive task, such as drafting a document or analysing a report, takes 30 to 60 minutes, so 30 to 60 dollars per task. The most expensive AI models today charge around 50 dollars per million output tokens. The same task produces one or two thousand tokens, which is 5 to 10 cents. That is a gap of several hundred times, at the highest price tier, and the price keeps falling.
The new engine also improves every few months. You can see the trajectory. Human intelligence stays where it is. Any owner shown a cheaper, faster engine on a steep improvement curve would start swapping engines the same day. The engine at the centre of your processes is being swapped whether you take part or not. Your only real choice is to move to the parts of your role that were never the engine: judgment, accountability, relationships, and taste. The engine was always the routine cognitive output.
This has happened before, and the people it happened to are glad
Every profession that went through this believed the routine work was the job. In each case the routine work turned out to be what kept people from the job.
Finance. When the spreadsheet arrived in 1979, calculation was the routine core of finance, and it was done by rooms of clerks. The spreadsheet made calculation almost free. Clerk jobs fell, on the order of hundreds of thousands in the US, while accountant and analyst jobs grew by even more. The people who liked the numbers moved up into work that valued their judgment. And cheap calculation created products that could not exist before. Michael Milken said the upheaval of 1980s finance came from spreadsheet software. Modelling that once took weeks by hand could be redone by changing one cell, and out of that came modern derivatives, a market now measured in the hundreds of trillions. The spreadsheet was one enabler among several, but the direction is clear: when a unit of financial calculation became almost free, the volume of financial work exploded, and a new professional layer grew on top of it.
Surgery. Before anaesthesia, a surgeon’s most prized skill was speed, because the patient was awake. The best surgeons could amputate a leg in under three minutes. Anaesthesia made that skill worthless overnight. The profession did not shrink. Freed from the clock, surgeons invented everything that needs hours of careful work: abdominal surgery, heart surgery, transplants. No surgeon alive wants the old skill back.
Banking. ATMs automated the cash-handling core of the teller’s job. Teller employment in the US rose for decades afterwards, because each branch needed fewer tellers, branches became cheaper to run, and banks opened more of them. The job rebuilt itself around relationships and advice.
In all three fields, the routine layer looked like the profession until it was automated. Then the real profession, which had been starved of time, grew into the space.
The three excuses
“My company doesn’t give us access.” Your employer’s software budget is not the ceiling on your skills. A personal subscription to a frontier model costs about the same as a streaming service. The skill of directing agents moves with you to every job you will ever have. Your company’s licence does not. Twenty dollars a month invested in the version of you that exists after this transition is the best return available to you, and it compounds into everything your career pays for, including your family’s life.
“I can’t put work data into an AI.” Correct, so don’t. Never enter your company’s name, your customers’ names, real financials, or anything identifiable, and follow your employer’s data policy without exception. This is not legal or compliance advice, and if you handle regulated data, ask before you paste. But none of that stops you practising. Abstract the details away. “A regional logistics company with three warehouses and a late-delivery problem” teaches you as much about directing an agent as the real client name would, because the skill you are building is how to structure and delegate work. And your job is not your whole life. Your finances, your travel plans, your side project, and your reading list are your own data, and you decide what to do with them. Practise on your own life until the skill is automatic.
“I don’t know where to start.” Have you looked? Have you asked an AI: “I am a [your role], I don’t know where to start with agents, build me a two-week plan”? That prompt works. You are reading an article about this right now, so the interest exists. What is missing is the first thirty minutes of trying. Pick the most repetitive thing you did this week and spend one evening handing it off. It will go badly the first time. That is normal, and it is still a start.
Start today
I am not telling you to be happy about the pace of this. The wind is uncomfortable, the rebuild is expensive, and the feelings are reasonable. But there is a difference between feeling the wind and fighting it. Fighting it burns the energy you need for the walk, and standing still burns time you do not have.
The clerks who moved up became analysts, and the surgeons freed from the clock built modern medicine. The people who lived through the last versions of this are glad it happened.
So answer the opening question honestly. If your working day looks the same as it did six months ago, you are standing still in the wind. Pick one task tonight and hand it off. Then another. Replan the route while there is still time to choose it.