"What should we actually be doing with AI" is a fair question. It's the one we get asked most. And it deserves a better answer than the noisy ones it usually gets.
Most of the advice in circulation isn't really an answer. It's a reaction. It pushes you towards one of two moves that look like opposites but come from the same place: a sense that something must be done, paired with no clear view of what.
The two reactive moves
The first is the small move. Get a few licences. Encourage people to use ChatGPT. Run a lunchtime session on prompting. The trouble is that vendors are switching on AI features faster than anyone can keep up, and following the features isn't the same as picking the work. The teams pulling ahead aren't chasing the release notes. They pick a few use cases that matter, get them working, and build capability alongside.
The second is the big move. A strategy refresh, a platform procurement, a transformation programme with a brand name on the cover. It looks bold. Boards approve it. Six months later the licences sit unused and the deck is shelfware, because AI capability isn't bought. It's built. The marketing functions getting ahead invest in the team they already have, not in software that promises to replace the thinking.
Both look decisive. Neither is deliberate.
What deliberate looks like
The deliberate alternative isn't a different action. It's a pause.
Before you choose a tool, choose a destination. Before you choose a destination, look honestly at where you are now. Pick the work, not the features. Most teams skip both steps and go straight to action, because action feels productive and thinking feels slow.
The thinking step is the one that pays off. It's also the one nobody is selling, because it doesn't come with a licence count.
Done properly, the answer to "what should we do" becomes concrete: your next step is one level up on each of your AI capability dimensions, on the way to a target you can articulate in two sentences. Not a moonshot. Not more ChatGPT. One level up, on each axis, towards somewhere specific.
Which raises the obvious question. One level up from where?
Two dimensions, not one
Most internal stocktakes grade AI capability as a single number. We're early days. We're getting somewhere. We're ahead of the curve. There is no such number.
There are two independent dimensions, and confusing them is why most stocktakes feel vague.
Knowledge & Advisory is how the team thinks about and applies AI. Do they understand where it fits in the work? Can they identify the right opportunities? Can they advise the rest of the business with credibility?
Tools & Delivery is what the team can actually build, deploy, and run.
Both matter. The order matters more. Teams that move forward deliberately tend to lead with Knowledge, because Knowledge is what makes the tool choices intelligent in the first place.
Five levels, plain English
Across both dimensions, the levels work the same way.
Novice. Limited exposure. A few people have tried things. No shared view of what AI does for this team.
Apprentice. Individuals are curious and experimenting. Useful for personal productivity. No shared approach, no embedded capability.
Practitioner. Confident and consistent. The team uses AI across the work lifecycle, and is building custom agents and knowledge bases to lift its own output.
Expert. Deep fluency. The team identifies where AI creates the most value and builds the agents and skills to prove it, advising stakeholders with credibility rather than hand-waving.
Master. Recognised authority. The team leads transformation, sets the pace, and contributes to the conversation outside the business. AI capability is part of what the organisation offers, not just how it operates.
You can't skip levels. Each one builds on the last. And the same gap looks very different over three months versus twelve, so the time horizon matters as much as the direction.
What the picture shows you
The point of placing yourself honestly isn't to feel ahead or behind. It's to see clearly what kind of move actually fits your situation.
The team that defaults to "let's all use ChatGPT more" is usually Apprentice on Knowledge and Novice on Tools. What they need first is a shared way of thinking about where AI fits in the work. The tools follow naturally once that exists.
The team that procures the big platform has been sold a Tools answer to a Knowledge problem. The strategy refresh and the platform deal do the work of looking decisive while skipping the thinking step. Their actual gap is wide on Knowledge, but they're spending as if it were on Tools. Six months later it shows.
The team that's Practitioner on Knowledge but Apprentice on Tools is in a different place again. They've done the thinking. They know where AI fits. They just can't yet build what their thinking says they should. That gap warrants different help, but at least the thinking has been done.
In each case, the right next step is invisible until you stop and look.
Faster is the start, not the finish
There's a temptation to judge AI by how much it speeds up the work you already do. Efficiency is real, and it's the right entry: start with what you do, do it faster, do it better. But faster is the start, not the finish.
The teams getting the most reach past efficiency into work that was previously too hard, too slow, or too unreliable to do well. The campaigns you wanted to run but never had time for. The personalisation you wanted to offer but couldn't operate. The analysis you wanted to commission but couldn't afford.
Ambitious, not chaotic. The work you wanted to do but never got to.
That work is only visible from the deliberate side of the pause. Reactive moves don't surface it. They're too busy keeping up with the noise.
How to look
You can do this on a whiteboard with the team. You can also do it in five minutes.
We've put the diagnostic online. Two dimensions, five levels, a personalised view of where you sit, a sensible target, and the shape of the gap between. It's a thinking tool, not a sales funnel. No sign-up, no follow-up call. You can find it at maturity-model.theforgedigital.com.au.
If the picture says the gap warrants a real plan, that's the conversation we have. We help organisations define an AI strategy and a roadmap to deliver it: where you are, where you're going, what gets you there, and in what order. If that sounds useful, get in touch.
The traits this work rewards are curiosity, judgement, patience, and optimism. They compound, and they're worth more than they were a year ago. You probably already have the people in the room. The diagnostic helps you see what to do with them next.