You don't have a position on AI yet. Neither does almost anyone else.

At every conference right now there's someone saying this changes everything and anyone standing still is finished. There's someone else saying it's overhyped, unreliable, and mostly a machine for producing confident-sounding rubbish. Both sound certain. Both cannot be right.

Somewhere in the middle is a leader with a team to run, wondering when exactly they were supposed to have worked this out.

Here's the first useful thing I can offer: your confusion is the most normal part of this picture. Feeling stuck isn't a reflection of your competence. The AI picture really is contested, it really is moving fast, and a great many people speaking confidently about it are also guessing. Nobody is as far ahead as the noise suggests. But there is a train leaving the station, and you need to find a way to get on it.

While the strategic debate runs, something more ordinary has been happening.

Various research findings suggest that upwards of 48% (as many as 80-something%) of people are using personal AI subscriptions at work, including inside organisations that have formally banned it.

This isn't defiance. It's people with too much to do finding a tool that helps, the same way other technologies have crept in over the years. They don't think of it as a policy question. They think of it as a way to get through their Tuesday.

Which changes the shape of the decision in front of you. You are not deciding whether AI enters your organisation. It's already in. What you're deciding is whether it's used thoughtfully, intentionally, and safely, with people knowing what's fine and what isn't, or quietly, on personal accounts, with your information in places nobody has vetted.

The shift worth making is from "we need an AI strategy" to "we need to say something useful about how we use this, and we can start this week."

There's a story running around that the organisations doing well with this started early, and that the window has closed. It doesn't hold up.

Almost nobody has a mature position on this yet. The tools have changed several times over in the last two years, so anything written eighteen months ago has to be rewritten anyway. The advantage sits with the organisations that started, not the ones that started first, and starting is a much lower bar than it sounds.

At the Revved summit in Auckland in August 2026, Frances Valintine put the choice plainly: use this with intention, or inherit someone else's version of the future. Her closing line was that both paths are hard, so pick your hard. Deciding nothing is also a decision. It's just one that gets made for you, slowly, by whoever is keenest that week.

You don't need to be a technologist to form a view. You need a small number of things you're clear on, and those are leadership questions you already know how to answer.

The order in Julia Pahina's framework matters. Start from the tool and you end up automating whatever is easiest to automate. Start from what must stay human and the rest gets much simpler. In most teams the list is short and obvious once someone says it out loud: how we tell someone their role is changing, how we handle a mistake, the conversation where somebody is struggling. Anything where the point of the interaction is that a person was on the other end of it.

The clearest proof of this isn't from a tech company. When IKEA rolled out its customer service bot, the obvious move was headcount reduction. They did the opposite. Over about two years they retrained roughly 8,500 call centre staff, many of them into interior design advisors helping customers plan a room rather than answer a delivery query. Remote sales grew 15 to 20% a year, revenue rose from around €1.08 billion to €1.25 billion, and their internal customer happiness score went from 60% to 89%.

The bot didn't produce those numbers. The bot cleared the repetitive queries, and then a decision was made about what the freed-up humans should do next. As Saba Samiei, who founded Macso Technologies, put it: "The steam engine didn't change the world. People who built with it did." The value came from the decision, not the technology.

So the test for any use of these tools is simple: did it hand somebody their attention back for the human parts of the work? If the hours saved quietly refill with more output, you'll have taken something that could have made your team more present or creative and used it to make them slightly more productive. That's a choice, and it's usually made by accident.

Take one thing you're using AI for, or thinking about, and ask yourself these five questions.

That last one is the honest test. If the answer is no, you're running a cost programme with a technology label on it. Fine, as long as everyone knows that's what it is.

You don't need to become an expert in any of these. You need to know they exist and know who in your organisation could find out.

Dr Alexandra Andhov, who chairs law and technology at the University of Auckland, urged leaders to look at their contracts and subscription terms, and to check whether they have the right to audit the AI they're using. It's the sort of question anyone would ask about a supplier and almost nobody asks about AI. You don't need to solve it this month. You just need to know who could find out.

Leading people through change is not nothing here. This is a leadership question wearing a technology costume, and you have answered its shape before: what are we doing, what won't we trade, who does this affect, and what is the smallest sensible first step.

Start where you are, with what's already true. That is enough to begin.