I often see people trying to figure out how they are supposed to work with AI. They look for the right prompt formula, the perfect template, the trick that unlocks better answers. Meanwhile, the simplest approach is right in front of them: talk to it, the way you would talk to a person.

That is how I work with AI today – like with a colleague or a friend I am getting to know. It has changed what I get out of it more than any technique.

None of this is new. You can read a lot about working with AI, and everyone puts together their own way of doing it. But I keep being asked how I actually do it, and things I take for granted turn out not to be obvious to everyone. So I am writing it down here – not as the right way, but as my way.

One clarification before I start: this article is about working with AI interactively – in a conversation with an AI assistant, where I lead every step. It is not about agents. In my series on the Agentic Operating Model, an agent pursues a goal on its own, decides within limits and acts in a process – which is why it needs an identity of its own and a clear place in governance. What I describe here is different: the assistant can carry out tasks for me, but only in our conversation, at my request and under my account.