I learnt the most about AI by keeping a journal

A year and a half ago, for three or four months, I kept a journal with AI every day. Sometimes I talked to it for fifteen minutes, sometimes for half an hour, sometimes longer. I kept coming back to matters that were genuinely on my mind.

The skill of working with AI grows through regular use in a real task.

When I mentioned it during a training session, I realised it had been one of the best ways of getting to know AI in practice. Not because a journal is some obligatory learning method. Because it gave me a reason to come back.

Why regularity teaches more than a demo

In a demonstration it is easy to be dazzled by the first answer. With daily use, further questions come up: whether the tool understood the situation correctly, whether its suggestion makes sense in my circumstances, whether I need to add context and whether I know what to do after the conversation.

The journal gave me plenty of such occasions. The topic did not end when I closed the chat. I still had my own matters to deal with, so I could compare what sounded good with what actually helped me sort them out.

I do not treat this experience as proof that AI is effective at solving mental health problems. I am describing my own practice of reflection and of learning a tool. Very personal information also calls for some thought about where it ends up. You do not need to share details of your private life just to learn how to work with AI.

How to apply it to a less personal task

The same mechanism works with a less personal task. For two weeks you can keep coming back to preparing lessons, learning a language or planning a project. What matters is that the task continues beyond the screen and that you can judge the answer on more than an impression.

I would start with a simple note after each attempt: what I wanted to achieve, what was missing, what I changed in the prompt and what I actually used. After a few rounds you build your own knowledge of working with the tool. Concrete, and tied to real needs.

In my case the recurring task was the journal. It helped me get my life in order and, along the way, gave me regular contact with AI. That is why I remember those months much better than many a feature overview I have watched.

What this means for a company

In a company you can apply this observation to learning a tool: instead of another feature demo, choose one recurring task, come back to it for a few weeks and note which answers turned out to be useful. This is a suggested exercise, not the result of an effectiveness study.