Intelligence has never been cheaper. Thinking has never been more expensive. Unfortunately, most organizations budget for the first and starve the second.
AI content has never been easier to spot, as you’re probably well aware. You glance at a deck, or read the first three lines of an email, and immediately know whether it was written by a person or by AI. This will only become more obvious as models improve.
The strange part is the inverse: when you read a piece of genuine human writing, the thinking in it almost glows. You can feel the angle, the choices, a human mind actually at work. This is the same instinct that lets you distinguish a handmade object from something mass-produced, even when you can’t explain how you know.
AI scales, and that is genuinely remarkable. But scaling something is not the same as using it well. What I want to get at here is the underlying part of all usage: learning to think deeply and then using AI to scale that thinking, rather than letting AI stand in for thinking altogether. That combination—thinking well and being fluent enough with the tools to multiply it—is going to be the differentiating skill in the coming decade’s job market. This issue impacts more than just individuals. Companies have to build the spaces, the training, and the habits that allow people to actually do it.
Walk into almost any company and ask where the line is between having too much and too little AI. No one will be able to tell you. There is no clear way to establish the baseline because the necessary vocabulary and policies don’t yet exist. However, people can usually spot the missing thinking in the first three lines. We detect the absence before we can identify it. Although the metric isn’t present, the reader’s instinct still registers its absence.
The tools you need to raise the baseline
Today, it’s easy to make something look good, but it’s hard to make it look excellent. The person who benefits the most from all this is almost never the one with the most prompts logged. Rather, it’s the person who thinks well and is fluent with the tools. Fluency multiplies no matter what thinking you put into it, but it never multiplies itself. Give the best tools in the world to a weak thinker, and you’ll get more weak thinking, just faster.
In response, the move most companies make is the wrong one. If you push AI usage, you get the same baseline running faster. If you raise the baseline, everything AI touches comes out sharper. But companies push usage anyway, because usage is what shows up on a dashboard.
Nobody’s baseline climbs just because a tool shows up on their laptop: someone has to build the conditions for it. Perhaps five to ten percent of any team starts with a high baseline, and if you give those people AI, they’ll pull ahead on day one. The other 90 percent is where the actual work lies. You have to create an environment where the whole team can raise its baseline, and you have to protect the existing thinking before the tools quietly erode it.
This shift is already in the data. Microsoft’s 2026 Work Trend Index, drawing on a privacy-preserving analysis of more than 100,000 commercial chats in Microsoft 365 Copilot, found that 49% of all usage sits in analyzing, reasoning, and deciding. That isn’t AI typing up documents. That’s AI doing the deciding. Inside that share, 28% is making decisions and solving problems independently, 19% interacting with others, 17% producing work, and 15% gathering information. If you still think of AI as a productivity layer that merely drafts your emails faster, those numbers should change your mind.
AI works with past data. Creativity, on the other hand, works toward something that doesn’t yet exist. You are the sum of every book, article, podcast, news story, and show you have ever consumed, but the reason people want to talk to you is because you aren’t a clone of any of them. You interpret those inputs in your own way and come up with angles that no one else would. People want your unique perspective, not a copy. That is the part of you that the model cannot reach.
The true development of a cognitive layer
Naming the layer was the first step. Opening the space was the second. Even with both in place, you still haven’t trained a single person. That is the third step, the one most companies skip because “training people to think” sounds like a category error. It isn’t. Thinking can be taught, and the most effective way to do so is also the cheapest method in this entire process.
Build a culture where people teach each other things often, as an ordinary part of the week. It barely matters what. Someone might teach how a transformer works, how to deglaze a pan, why their last project went sideways, or the offside rule. What matters is the act of teaching, because it engages all four “thinking muscles” at once:
- Listening and understanding
- Thinking and extrapolating
- Challenging and validating
- Applying and learning
That is what makes it the best drill available. To teach anything, you first have to listen to and understand the learner’s confusion before offering a solution, or you end up answering a question nobody asked. You have to think and extrapolate, pulling ideas out of your head and shaping them in a way that can be transferred to whatever medium the learner needs, even if you had not prepared for it. You have to challenge and validate on the spot, holding the idea up against questions you didn’t see coming. And you have to apply and learn, rebuilding the explanation on the fly when the first version falls flat, which it usually does.
Four muscles, one act. Teaching is the one exercise that works all of them at the same time.
This is more than a hunch. Educational psychology calls it the protégé effect: people learn material more deeply when they prepare to teach it and then actually teach it to someone. That is what Fiorella and Mayer pulled together in their 2016 review of generative learning strategies. Teaching is not a tax you pay on top of learning. It is one of the most efficient forms learning takes.
This leaves the leader with one important job: creating opportunities. Weekly lightning talks carry most of the load. Ten minutes, teach the room one thing. Anybody can lead, and the subject can be anything. The rotation is tight enough that nobody can quietly slip out of it. Reverse mentoring handles most of the rest because, the moment a junior person teaches a senior person, the normal hierarchy that keeps honest questions unasked flips, and both have to adapt.
The remainder is just “seasoning”: teach-me-one-thing pairings for the gaps in the calendar; five-minute demos from those who shipped a project; knowledge cafés when a subject warrants more time. The cheapest curriculum a company can implement is the standing requirement that its people often teach each other.
Offer space to “waste thinking time” productively
Thinking is never a waste, no matter how it looks from the outside. It sounds banal to say that it’s a muscle you train, but I have no better analogy. Carrying weights at the gym or carrying your groceries up the stairs strengthens the same muscles.
This is something companies don’t understand, and even when they do, they execute it poorly. It is genuinely hard to manage. How do you give people room to train without locking your resources in eight-hour workshops just to tick a box on your learning and development roadmap?
Start by setting aside time. Time is the most important asset for everything. Not only for active learning, but for resting cognitive capacity, which directly impacts the quality of thinking. Just as we set aside time for people to eat lunch, a post-industrial company must set aside time for people to “waste on thinking” to improve its overall quality.
The real question is how to waste time productively. One way is what I mentioned earlier: teaching sessions on a work-related or non-work-related topic. Preparing to teach is training in itself. Another way is to run open debates and discussion forums with no required output. This works best in small groups of three or four around a question such as, “Are we even targeting the right audience?” The question must be concrete and sharp, or the debate will never get off the ground. Even if nothing actionable results from it that week, it pays off in the long run.
Giving people space to think outside the operational grind is essential if you want a cognitive layer that compounds. Companies that understand this and don’t chase the next quarter’s productivity bump will retain the highest-quality thinking at the enterprise level.
None of this is easy. It takes time and effort, and that’s exactly the point.
Thinking in the age of AI is hard, genuinely hard, not because we have forgotten how, but because when you can produce anything at the push of a button, sitting down to actually try feels wildly unproductive. Of course, it’s easier to delegate. But as we said at the start, it’s painfully easy to spot when AI was used and no one was thinking. Intelligence has become cheap. Thinking is the most valuable asset a company can have right now, yet almost no one is investing in it.
