I have used artificial intelligence tools every single day for almost two years. I started out of curiosity in December 2022 and kept going because they became part of my routine, like my phone or my inbox. Along the way I have written here about how I use ChatGPT in business, about what small companies can automate, and about agents. This time I want to write about something else.
Because, looking back, the most important lessons are not about the tools. They are about me and the way I work. AI has acted like a mirror: it showed me what I do well, what I do out of habit, and what I was doing just to feel busy.
1. AI showed me what is thinking and what is just busywork
The first surprise was how much of my daily work a machine could do decently. Summaries, first drafts of emails, presentation outlines, rewording, translations. Things that filled my day and gave me the feeling of being productive.
The second surprise was what it could not do. Decide whether I should join a startup. Sense when a founder is telling the truth and when they are selling a dream. Choose between two good options. That was the real work. Most of the rest was busywork dressed up as work. Not a pleasant discovery, but a useful one.
2. The bottleneck is no longer the first draft, it is the decision
When a first draft arrives in seconds, the problem moves. You are no longer staring at a blank page. You are staring at five versions, and you need to know what you want. If you do not, AI will not help you: it will just give you more versions.
I realised that the time I saved on writing I was losing to indecision. So I started asking the question before, not after: what decision do I need to make, and what information am I missing? Most of the time, answering that does half the job.
3. A good prompt is a good brief. And that made me a better manager
At first I wrote vague requests and got vague answers. Then I learned to give context: who the audience is, what the goal is, what to avoid, what a good result looks like. The answers improved immediately.
The interesting part is that the same discipline carried over to working with people. If I have to explain clearly to a model what I want, why would I expect the people on the teams I work with to guess? Many of the frustrations I had as an entrepreneur and advisor came from poor briefs. Mine, not theirs.
4. The trap: more instead of better
This is what I see most often around me. When you can produce anything in seconds, you produce too much. More emails, and longer ones. More documents nobody reads. More posts saying the same thing. The noise grows, but other people’s attention does not.
I now have a simple rule: if AI helped me write something, I try to make it shorter, not longer. A five-line email that says clearly what you want is worth more than a full page, however well phrased.
5. Thinking time must be protected, not outsourced
There is a temptation to let the model think for you. You ask the question, you get a convincing answer, you move on. It is comfortable, and it is dangerous. Judgement is a muscle: if you do not use it, it weakens.
So I keep blocks of time with no screens and no assistant, when I just think. About strategy, about people, about what is not working. AI comes afterwards, to check, to add, or to push back. Not before. Two years ago I wrote about burnout in the online generation, and the lesson still holds: when everything gets faster, breaks become more valuable.
6. I verify everything
The models have improved a lot. OpenAI’s o1 reasoning models, released in September, solve problems that a year ago seemed out of reach for a chatbot. Just yesterday, ChatGPT search launched, complete with sources. And yet they still make mistakes. Confidently, in polished sentences, and sometimes with made-up numbers.
My rule is simple: any figure, name, date or quote that leaves my desk must be checked at the source. If I do not have time to check it, I do not use it. A reputation takes years to build and one email to damage.
7. The best sparring partner I have, at any hour
The most valuable use of AI for me is not writing. It is disagreement. I describe an idea and ask it to tell me why it would fail. To play a sceptical investor, an unhappy customer or a competitor. It often raises objections I would only have thought of a month later.
ChatGPT’s Advanced Voice Mode, rolled out in September, makes this even more natural: you can think out loud during a walk along the seafront, no keyboard needed. It does not replace a good friend or mentor, but it is available at two in the morning and does not take offence when you argue back.
8. Energy and attention matter more than tools
Something new comes out every week. Audio overviews in Google’s NotebookLM, Anthropic’s “computer use”, the agents I wrote about this summer when I argued that AI agents are the next step. If you try all of them, you stop working and just keep testing.
I have learned that the best tool is the one you use well, not the newest one. And that a day when I slept well, exercised and had a clear head is more productive than any subscription. Tools amplify who you are. If you are tired and scattered, they amplify the tiredness and the scatter.
9. Life is not a workflow to optimise
I travel a lot, I live in Cyprus, and I have a family and projects in several countries. AI has helped me work more efficiently from airports and hotels. But it has also tempted me to work more, simply because I can.
I have come to the conclusion that saved time does not have to be reinvested in work. Sometimes the best thing you can do with an hour you have gained is to go out with the kids, play tennis, or do nothing at all. Productivity that eats your life is not productivity. It is just another way of getting lost.
The habits I kept
- Before opening an AI assistant, I write down in one sentence what I want to get out of it.
- I give context the way I would in a brief: audience, goal, tone, what to avoid.
- I shorten everything I write with AI’s help before I send it.
- I check any figure, name or quote at the source.
- I ask the model to argue against my ideas before I present them to anyone.
- I keep screen-free thinking time every week.
- I do not make decisions about people or money based on a generated answer.
If I had to sum it up in two sentences: AI has not made me smarter. It has forced me to be clearer. For entrepreneurs and for small businesses looking to automate, I think that is the real lesson.
What have you learned about yourselves since you started using AI? Has it made you more productive, or just busier?