I am a surgeon. In my own field, where I am already trained, AI has added very little. The gain was small because I was already good at the thing it was helping me with.

That is the pattern in the research, not just my experience. When experts and beginners both start using AI, the experts gain little, and the beginners gain a lot. In the study of consultants that Ethan Mollick worked on, the lowest performers improved by 43 percent, while the top performers improved by 17 percent. In a study of more than 5,000 customer service staff, the least experienced gained 34 percent, while the most experienced gained almost nothing. AI helps the people who know the least, far more than the people who know the most.

So I am the wrong person to tell you what AI did for me. I had the least to gain. Being that expert is what showed me who has the most to gain. They are the people I treat every day, and they are nowhere near this conversation. If those of us who understand AI do not bring them in, the tool that could close the gap will widen it.

I have spent years treating patients and trying to help them become their own advocates. The hardest part was never the treatment. It was prevention. Most chronic diseases are preventable, yet we are not good at preventing them.

We hand people statistics, infographics and government campaigns. None of it speaks their language, and none of it reaches them where the decisions are actually made: in the supermarket aisle, at the menu, in their own kitchen. I can tell someone to do their pelvic floor exercises, but I might not see them again for six or twelve months, and by then my advice is gone. Their primary care doctor can talk about salt, sugar, alcohol and exercise in a ten-minute appointment that fades by the weekend.

AI does the thing we cannot: be there in the moment of the decision. You can photograph the menu before you order. You can ask what to cook with what is in the fridge, for less money and less waste. You can stand in the aisle and check which product fits your goals. It is a tutor who never closes and meets people in their own lives, rather than in my consulting room once a year.

I am a doctor, and it still taught me. I thought Bircher muesli was the healthy choice for breakfast, but AI told me to have poached eggs, because I am trying to maintain muscle after menopause. If it can correct me, think about what it does for someone with less health education.

In medicine, a treatment that works in a trial but never reaches the patient is a lost opportunity, and AI is full of opportunities that are not reaching the people who need them most.

So why are these people not already using it? Not because they cannot. The barrier is rarely capability.

For some, it is access, or never being shown how. But the deeper barrier is fear. AI is complex and unfamiliar, and most people meet the unfamiliar with worry rather than curiosity. They have no technical background, so the whole thing feels like it belongs to someone else. And the loudest voices they hear are catastrophic: AI will take your job, your privacy, your mind. Fear is easy to sell. It is far easier to frighten someone about AI than to show them what it can do for them.

The numbers bear this out. More than a third of people still do not use these tools, and the gap is widest exactly where you would expect: by age, by education, by income. Around half of non-users say they do not know how to use AI well, and most simply do not see what’s in it for them. These are not stubborn people. They are people no one has brought along.

Fear is the enemy of the curiosity and courage it takes to benefit from AI. And the people most ruled by that fear are the same people with the most to gain.

The obvious objection is that this fixes itself. The tools get cheaper and easier every month, so the people on the outside will drift in on their own, and the gap closes without anyone doing anything.

I do not think it works that way, and the evidence is starting to agree. When people use AI unaided, the ones who are already capable get more out of it. They question what it tells them, they catch it when it is wrong, and they push it harder. People who are uncertain do the opposite. They ask less, they verify less, and they take what it says at face value. A study of knowledge workers found exactly this: the people with the least confidence in their own judgement did the least checking of the machine. Access on its own does not level the field.

There is a second catch. Levelling shows up in the studies at the level of a single task, the novice who suddenly performs like an expert. It does not show up at the level of the economy. Gains are being concentrated in a handful of firms and countries, and the distance between those ahead and those behind is growing, not shrinking.

Left alone, the same tool that could close the gap is already widening it. If we want AI to be a great leveller, someone has to do the levelling. The two things sit together. The prize is largest for people at the bottom, but it is also hardest for them to reach on their own.

There is a risk in this, and it is worth naming. Hand someone a tool that always sounds sure of itself, and they can lean on it without ever checking it. The people most likely to do that are the unsure ones, the people who cannot yet tell when it is wrong. Used that way, AI does not lift them. It makes them dependent. The more someone trusts it, the less they tend to question it.

So bringing people along is not the same as handing over access. The part that matters is judgement: how to question an answer and how to tell when it is wrong. That is the skill worth teaching, and it is why the guardrails and the accountability still sit with the people who build and run these tools.

If you are reading this, you are on the inside. You already use AI, and like me, you are near the ceiling of what it will do for you. That is why you are the person who can close the gap. You can see what it does, and you are not afraid of it.

The tools already work. They speak plain language, and they are mostly free. Nothing new has to be invented for the people on the outside. What is missing is not a better model. It is someone to show them the one that is already here.

Using AI well is its own skill: the back-and-forth and the habit of questioning what it tells you. It comes naturally to you, so it is easy to take for granted. Most people have never been shown it. It can be taught, and you are the one who can teach it.

And the payoff is not only theirs. More people using AI well takes pressure off the things we all rely on, like overcrowded hospitals, where staff could move from putting out fires to preventing them.

So if your hours are going into using AI to get further ahead, spend some of them bringing other people with you. Sit with one person and show them something real. How to put a confusing bill or a phone contract into plain words before they sign. How to check what a medication is for. How to draft the email they have been dreading. The ones who need it most are not in your tech hubs, your AI meetups, your online chat rooms. They are everywhere else.

There is also a reason this cannot wait. A head start in AI compounds. The people already using it get better at it every month, and that skill builds on itself. The longer someone stays outside, the further they have to travel to catch up. There comes a point when the divide is too large to cross, so the time to close it is while it is still small enough.

So what do these people actually need? Three things. Access to the tools. Someone to show them how to use them. And a sense of where and when the tools can help.

None of that requires a breakthrough. It requires the people who already have all three to hand them on.

AI could be the great leveller. The thing that finally gives the person with the least the same standing as the person with the most. Or it becomes the great divider, the fault line that decides who gets ahead and who gets left behind for good. AI cannot do this. We choose who we bring with us.

The AI Collective is built by volunteers across 180+ chapters in 40 countries.

Thank you to the thousands of volunteers around the world who make this work possible. We truly could not do this without you.