I run my business with a team of AI colleagues. Here is what they taught me about people.

Penny, the chief of staff, runs contacts, follow-ups and meeting prep. Wells, the copy chief, argues with my headlines. Ada, the web director, keeps this site honest. Ogilvy shapes the offers and proposals, Nash preps me for the hard conversations, Perkins edits anything in my author voice, Sterling reads the numbers and keeps the company secretary happy, and Brunel designs the programs we deliver. Eight colleagues. None of them is a person. Each has a name, a charter, a memory and a log, and none of them touches anything live without my yes. (Ada built this page. Wells wrote it. I edited it, which they both expect.)

I say this because it is the honest answer to "so what do you know about AI?" I know what it is like to run on it every day, with real clients and real money at stake. That is a different kind of knowing from reading about it, and it tends to be the kind leaders ask me for.

The first week was humbling. I asked for things and got back fluent, confident, slightly wrong answers, and the fault was mine every time. I had not said who the reader was, or what good looked like, or what to leave out. I had been teaching that discipline to sales teams for years and skipped it myself, because the machine was so quick to oblige.

Briefing an AI is a communication problem. Who it should be, what it needs to know, exactly what to do, what the result should look like. The same four things that make a board paper land. Once I treated the machine as the most literal colleague I had ever met, the quality of the work went up and the drama went down.

The second lesson arrived at about 2am a few weeks in. If it can do the first draft of everything, what am I for? Most people I teach have met that thought and said nothing about it, which is why so much AI use at work happens in private. The way out is to stop asking the machine to be faster and start asking it to make you better. Faster is a shrug. Better is a different job. New, the things you never had time to attempt, is where it gets interesting.

I want to be better, not replaced. Someone says that in every room, once they feel safe to say anything. It is the most useful sentence in the whole subject.

Three things, in order.

Teach people to think before they prompt. A method, practised on real work, with the guardrails for your sector built in. Most of the gap between organisations that get value from AI and those that do not tends to sit here, and it is the cheap end of the problem.

Decide what AI is for in your organisation before someone decides it for you. Which work goes in, which stays out, what you will measure, who is accountable, and how you will encourage people rather than police them. Six decisions, made in the right order, by the people who own the outcome.

Say the human part out loud. Tell people you want them better, not replaced, and then build the incentives so it stays true. The technology is the easy part. The psychology is where the innovation is.

I teach Practical AI for Leaders and Professionals, a three-quarter day that changes how people think with AI (93% participant satisfaction across six NSW public sector cohorts this year). I run a program for executive teams who need to make those six decisions properly. And I give a keynote called Most people are using AI too small, which is the talk I most enjoy giving.

If any of that sounds like your organisation, send me the brief. I will say plainly what fits and what does not.