In 1994 my mother stopped working. She was about 36.
The Soviet Union had ended three years earlier, the economy she had trained for was disintegrating, and she went home to hold a family together while everything around us was rebuilt from nothing. That was work 24/7 without salary and nobody called it a career.
We perceived it as a normal, many families around us lived the same. My mom never returned to work. I think she most likely thought about that but the gap between her skills and the fast evolving world became bigger every year.
There were no evening courses. There was no internet full of teachers waiting for her. There was no version of 1994 in which a woman with children and no income could quietly retrain herself between other obligations.
My father is 68 and still working. He has his own business renting out real estate, and he runs it much as he has run it for decades. He built something that worked, and he was never given a reason to change it.
The world changed anyway.
I am 46. It still surprises me to write that number. Somewhere in my head, 46 means what it meant when I watched my parents live it: settled, decided, already grandparents.
For a long time I thought the important question was: when does a career end? Watching my parents, I no longer think that is the question. The real one is: when does the learning quietly stop?
My mother’s moment came at 36. Not at 65. And nobody announced it, including her.
The door does not close at a safe, distant age you can prepare for. It closes whenever you stop, and it makes no sound.
I started because I was frightened
This is a letter about learning AI.
I suddenly saw that the world, technology has evolved so much without my even realising that.
Two years ago I noticed something I did not like. I was delivering results, I was busy, I was good at my work, and the way I worked had not meaningfully changed in a long time. I was using the same methods I had used for years, and telling myself I would learn something new when there was time.
Why did I bring my parents’ story into it? Because I recognised the pattern. It was my father’s business and my mother’s 32 years, arriving in a more educated form.
I did not start learning AI because I love technology. I started because I was frightened.
And look at what I actually did about it. I decided I needed to become a data scientist. I applied to a master’s programme in the United States and was rejected, because my background was not IT or data science. Then I spent a year looking for courses in Python, SQL and statistics. Expensive ones, demanding months of evenings I did not have.
At Easter this year I stopped preparing and built something small and ugly with AI instead. Four months later, everything about my working life is different.
What I would do differently
If the first half of this letter is the warning, this half is the way through the door. 7 things that would have saved me the year I lost.
First, the honest part: I did this the expensive way. 3 hours a day, almost every day, and my weekends. I am not proud of it and you should not copy it. That is precisely why I am writing this. I spent the hours so that you do not have to.
1. Do not try to become technical first.
You do not need Python, SQL or a qualification in data science to use AI. I lost the best part of a year believing otherwise. You need to build one small thing that solves a problem you actually have.
2. Start with what makes your own life difficult.
Not a tutorial. Not a course project. The thing that genuinely irritates you.
Mine was German. I had been trying to progress for a long time and I was never satisfied with my level. So the first thing I built was almost embarrassing: a Google Sheet that gave me an English sentence, let me type the German translation, then marked it and gave me the pronunciation. That was all it was.
When I connected the keys that let one program talk to another, and a basic screen appeared, and it actually worked, I wanted to jump. Nothing in those four months taught me as much as that ugly little tool.
3. Stop buying courses. For someone like us, the AI itself is a better teacher.
This is the thing I abandoned, and I do not regret it. A course is built for a general student. It cannot know that I have more than two decades in commercial roles, that I think in customers and margins, and that I have no technical vocabulary at all.
I can simply ask Claude, and get the explanation built around what I already know, at my level, in my language, with nobody watching me not understand. If you have ever felt too old or too slow inside a beginner tutorial written for a 22-year-old, that is the whole difference.
4. Brief the AI the way you would brief a capable colleague.
My commercial experience became useful when I stopped treating AI like a search box. “Summarise this” gives it very little. “Turn these meeting notes into a one-page decision brief for a senior commercial leader, separate decisions already made from open questions, do not invent missing information, end with owners and next actions in a table” gets work I can use.
5. Separate the thinking from the building.
My most expensive mistake. I was thinking with Claude and building in parallel, working out what I actually wanted while it was already producing. So it built the wrong thing, I corrected it, it built the wrong thing again, and I burned through my weekly limits in days.
Now I think first, with it, until I know exactly what I want. Only then we build. It costs a fraction of the money and the result is better. Nobody told me this, because everyone technical already assumes it.
6. Match the checking to the risk.
AI can produce a confident answer that is incomplete or simply wrong. So from the first day, I would match the level of review to the level of risk: a private checklist needs less scrutiny than anything touching a customer, money or another person. My simplest rule: AI may prepare. I judge and approve.
7. Apply it in your job: automate a small piece, or build something useful.
This is the part that changed things for me, and the part most people will resist.
I saw that the business needed better market and performance data, so I offered, on top of everything I already had to do, to build a regional dashboard. A few weeks later there was a working interactive dashboard. The business uses it today. Nobody assigned it to me. There was no budget and no mandate.
If you wait for your company to train you, you will wait. If you solve one real problem where people can see it, everything changes.
If you tried this before and hated it
When I started, the tools were fragile. The model could destroy a whole project in minutes and I would spend hours putting it back together. That is not the experience anymore. They improved enormously in a matter of months. If you formed your opinion a year ago, your opinion is out of date.
The question that stays
I intend to work another 15 or 20 years. I have three children and I want them to start their lives from firmer ground than I did.
But my mother’s window closed while she was busy doing something necessary. That is what stays with me.
So the question I would leave you with is not: is it too late for me. That question has no useful answer, and asking it is mostly a way of avoiding the next one.
Ask this instead: when was the last time I learned something that changed how I work?
I could not have answered that honestly two years ago. If your answer is measured in years, the door is already moving, quietly, the way it did for my mother, while everything else was demanding your attention.