the ai mindset

In the companies where I set up and run the AI, I keep seeing the same two people at the same screen. Same model, same afternoon. The first one types in his question, drops the answer somewhere, and on with his day. Done. The other one? Stares at that same empty prompt field like it’s a blank sheet of paper. Hands the AI the real situation, argues with what comes back, bins the first three tries and keeps hammering on the fourth. Right until there’s something standing that he could’ve thought up himself. Only faster.

Same cost, right? Same bill at the end of the month. And still the two run so far apart you almost don’t believe it. It isn’t the machine. It’s what those two brought with them when they sat down. That’s the AI mindset, and it sits in two places at once. The user brings the creativity and the judgment. But the owner hangs the ceiling: what the AI’s allowed to do, which systems it gets into, what infrastructure’s there. And under that ceiling the user makes whatever fits. No more than that.

the small ones are ahead

And here’s something that surprises people. Of the small businesses I meet, pretty much 100 percent are already way ahead on this mindset. Not because they’ve got big budgets, the opposite. It’s because their owners are driven, and they’re often already deep into AI themselves, simply because they haven’t got many people. Cost and output are everything to them. They haven’t got all that money lying around, so getting more out of every hour matters to them in a way it just doesn’t inside a big enterprise.

And the big enterprise? Much slower. Security weighs heavier there, and fair enough, they handle far more data and they’ve got a media presence to protect. So there, the AI mindset really is a top-down thing: it takes the right people, with the right mindset, making the right calls on where that ceiling hangs. Get it wrong at the top and it barely matters how sharp your people are underneath.

ask-and-hope, and the trust you lose

Because using AI in this era as a pure chat box or a search bar? That’s underselling it, badly. It misses what it can do to your output, to the quality of that output, to your efficiency, more than you’d put in numbers. Supercharge every employee with the right tools and it’s genuinely game-changing for any company. But that starts with the mindset at the top. If the owner only ever uses the chat box, so will everyone else, because nobody’s handed them the tools to hook agentic AI onto the systems they actually work in.

And that ask-and-hope habit has a price you don’t spot right away. There’s good research on it: hand people the exact same AI work but tell one group it came from an “AI coworker” and they catch fewer of its mistakes and push less of it up to the manager. The vigilance drains away, and that vigilance is the actual job.

But here’s what I see go wrong most, over and over. People reach for the wrong model, for the wrong kind of task, with the wrong context. They open a chat window to wrestle an Excel sheet or a PDF, and of course the output’s rubbish. And then they lose trust in the whole thing. “My model can’t even do this. It’s always wrong. Look at this PDF, it looks like garbage.” But that’s not the model failing. That’s someone who isn’t IT-minded, who won’t try five models and five different ways, and honestly often can’t, because they haven’t even got the tools to reach that software. Know the right model and the right way to use it, and you get the best out of it. I’ve made PDFs, motion graphics, whole documents I never imagined I could make. Right tools, right MCP surface, right models.

hands and memory are a choice made up top

Straight out of the box, an AI is missing two things, and both get decided above the user’s head.

Hands first. It isn’t wired into the systems where the actual work happens. Does it get access to those? That’s a board decision, about integrations and security. Nothing a user just taps into a prompt. And memory second. It plain doesn’t know your organization. Paste in a PDF and you’ve handed it a snapshot, not a company. Whether the AI has real, well-chosen context to work from hangs entirely on how the infrastructure got laid down. Top again.

And watch what an AI does to a process, because it works like a mirror. Point it at a messy process and you just scale the mess. It won’t paper over anything the way a person still would. Which processes you plug in, and whether they’re halfway in order, that’s one more thing that lands at the top.

what opens up when the top builds

Flip it round, and everything opens. An owner who dares to put the infrastructure in place, responsibly: the AI gets into the systems that matter, gets real context instead of a document dump, gets clear boundaries to act inside. That’s the high ceiling. And don’t let anyone tell you AI’s still in its early days, that it can’t really do much yet. That myth’s just wrong, and I say that from first-hand experience of what you can actually build. Organizations wildly underestimate what’s possible. Of course you’ve got to be careful, of course you need the right setup, but it’s far from impossible. It’s very doable, and safe.

So where do you start? With your IT branch, not your sales floor. Don’t trust your sales or your service people with it yet? Fine, but get the base layer right in IT first. No IT branch at all? Then bring someone in from outside who sets it up, so you and your people get the tools to use agentic AI responsibly. Because it starts with mindset, top-down, and it ends on the shop floor, with retraining people to actually create with AI. Not just ask a question and get a block of text back, but build with it, weave it into every work process. Get that base layer down and you’re ready for anything. After that, you only really need critical thinking.

the skill you protect

The teams that get real value out of it? They’re rarely running the newest model, or the slickest partnership with their digital colleague. They’re just the ones where the top dared something and the people stayed sharp. Two mindsets that have to click, or nothing happens.

So the skill worth protecting isn’t the one called AI literacy. Up top it’s nerve, plus the clarity to build the right infrastructure. On the shop floor it’s creativity and judgment. And that first bit, the top-down bit, is exactly where owners get stuck and come looking for help. I wrote a book about it, and getting those calls at the top right is what I do.