I use AI to run my business every day, and it’s never once given me a finished answer

If you look at my week, AI touches almost everything I build at NITO. I use it consistently for positioning, competitor research, marketing plans, and analysing cohort data.

And yet two years of using it daily has taught me that AI has never once given me a finished version. Every output has always been a starting point, a draft that’s maybe 60-70% of the way there, shaped entirely by the context I give and how hard I push back on its responses. Most people treat this as writing a good prompt and walking away. It's a constant loop with four main steps.

First, you give it real context; not a one-line brief but your actual positioning documents, your own writing, data, and real numbers. If you don’t know what you want, you will get a vague response that sounds good but is too generalized to work for you. Second, you train it and correct its assumptions early, before it builds an entire answer on a premise that’s wrong.

Third, you challenge it. You push back on its reasoning and make it defend its logic, as you would do with a sharp analyst on the team. If it can defend its answer, that’s when you start building on it.

Fourth, you finish it in your own words. AI can help you structure it, but it still needs your judgement on top. The version you can ship and build upon is never the raw output.

This is where you build a complete version through context, training, challenging, iterating, and finishing - and this complete loop is a real skill, a human one. The same judgement, subject-matter expertise, and critical thinking that made someone good at their job before AI existed are exactly what make them good at directing AI now. Someone without that judgement doesn't get worse results occasionally; they get confidently wrong results, at speed, every time.

This is where most people either start using AI extensively or completely stop using it. It’s not a talent you either have or don't. It's a skill you build the same way you'd build any other, through training, repetition, and feedback over time. I wasn’t great when I first started using AI to build NITO. I'm better at it now because I've done it hundreds of times, and there’s no quick prompt to help with it.