When Intelligence Becomes Abundant, Focus Becomes An Advantage

There is something I have been thinking about as I watch people work with AI.

We now have access to an extraordinary amount of intelligence on demand.

A developer can understand an unfamiliar technology in minutes. An employee can research a subject they have never worked on. An entrepreneur can study a new industry, analyze competitors, create a business plan and start building a product in a fraction of the time all of this would have taken earlier.

This is a remarkable advantage.

But I increasingly think it creates another consequence.

As access to intelligence becomes increasingly abundant, our own attention becomes more valuable.

AI can give us breadth almost instantly.

Focus is how we build depth.

Before AI, entering a new domain had considerable friction.

You had to search, read, speak to people, understand terminology and gradually piece things together.

Today, you can have a reasonably intelligent conversation about an unfamiliar subject almost immediately.

I use this capability myself. It is enormously useful.

But there is an important difference between having access to knowledge about something and understanding that thing deeply.

AI can explain a software architecture to you.

That doesn't mean you understand why your particular product ended up with its architecture.

AI can analyze an industry.

That doesn't mean you understand how customers in that industry actually behave.

AI can suggest what a business should measure.

That doesn't mean you know which number inside your business is telling you something important and which one is simply noise.

Those differences come from context.

And context takes time.

Think about someone who has worked deeply on the same product for several years.

They don't simply know the technology.

They know why certain decisions were made. They remember approaches that looked promising but failed. They know which customer complaints keep appearing. They know which apparently small changes have consequences elsewhere.

They begin to recognize patterns.

A new person can read the documentation. AI can summarize the entire codebase. Both can become productive much faster than they could before.

But neither instantly inherits all the context accumulated by someone who has spent years thinking about the product.

This distinction matters to me.

AI has enormous knowledge. Humans can still have unusually valuable context.

And focused work is one of the ways we accumulate it.

This is one of the risks I see in how people use AI.

You can now produce useful work in a domain before you understand that domain particularly well.

That is often a good thing.

Someone learning a new technology can become productive sooner. A junior developer can solve problems that previously required more experienced help. A business person can analyze data without becoming a data scientist.

Large field experiments involving 4,867 software developers at Microsoft, Accenture and another Fortune 100 company found that access to an AI coding assistant increased completed tasks by 26.08% overall, with less-experienced developers showing higher adoption and greater productivity gains.[1]

But productivity and understanding are not the same thing.

1. You can generate more code without becoming a better software architect.

2. You can produce more analysis without developing better judgement.

3. You can research more industries without understanding any one of them deeply.

AI can accelerate your output faster than it accelerates your understanding.

My concern is what happens when we confuse the two.

I have always seen differences between people who stay deeply involved with something and people whose attention constantly moves.

AI makes me think that difference may become more important.

If someone remains at the surface of many areas, AI can help them enormously. But it can also help almost everyone else operate at that surface.

A reasonable explanation, first draft, initial analysis or piece of code is becoming easier to produce.

So where does an individual advantage come from?

I increasingly think part of it will come from going further down.

1. Understanding the customer better.

2. Understanding the product better.

3. Understanding the technology beyond the immediate task.

4. Knowing the history behind decisions.

7. Recognizing exceptions.

8. Knowing when the obvious answer doesn't fit.

That kind of understanding is difficult to accumulate when attention is constantly being reset.

Research on what Sophie Leroy calls attention residue supports at least part of this problem. Her experiments found that when people switch away from unfinished work, part of their attention can remain with the previous task and performance on the subsequent task can suffer.[2]

That research predates today's AI tools, but I think the underlying problem becomes even more relevant when technology makes it so easy to move between ideas, tasks and domains.

This is why I suspect people with scattered attention may find it harder to differentiate themselves in an AI-enabled workplace not because AI makes them incapable, but because AI makes shallow capability much easier for everybody to obtain.

One assumption about AI is that domain knowledge will matter less because AI already knows so much.

I increasingly believe the opposite may happen.

Research on workforce reskilling for AI points in a similar direction. Prasanna Tambe's work in Management Science finds that AI and algorithmic capabilities act as complements to domain expertise, with greater value created when domain experts themselves can interpret and apply these tools.[3]

The more capable AI becomes, the more useful it becomes to have someone who knows what to ask of it, what to ignore and when its answer doesn't fit reality.

A person with deep domain knowledge can use AI differently.

1. They can challenge assumptions.

2. They can recognize missing context.

3 .They can notice when a technically correct answer is commercially wrong.

4. They can connect an answer to things that happened months or years earlier.

5. They can distinguish between something that sounds intelligent and something that matters.

This is not a reason to compete with AI on how much information we can remember.

That competition doesn't make much sense.

It is a reason to develop enough depth that all the intelligence available to us can be applied with judgement. This is why communication is becoming an execution skill in the AI era, because applying that judgement effectively depends entirely on how clearly you direct the AI.

I see another version of this problem for entrepreneurs.

AI has dramatically reduced the friction involved in pursuing an idea.

Have another business idea?

- Research the market.

- Ask AI to analyze competitors.

- Create the positioning.

- Build a prototype.

- Write the landing page.

- Start another experiment.

A lot of this can now happen incredibly quickly.

Again, that is a wonderful capability.

But I think there is a hidden risk.

AI can make scattered attention look like productivity.

An entrepreneur can be extremely busy researching five opportunities, prototyping three products and planning two new businesses.

There can be an enormous amount of output.

But output doesn't necessarily mean that enough time has been spent understanding one customer, one problem or one market deeply enough to discover something meaningful.

AI has reduced the cost of starting things.

It hasn't reduced the value of staying with something.

Perhaps it has made that discipline even more important.

Quite the opposite.

I think we should use AI aggressively.

1. Use it to research faster.

2. Use it to code faster.

3. Use it to analyze more information.

4. Use it to challenge your thinking.

5. Use it to remove work that doesn't deserve your attention.

But then use the attention you have saved deliberately.

If AI saves you 3 hours, the biggest opportunity may not be to fill those 3 hours with 20 more shallow tasks.

It may be to spend some of that time going deeper into the work that actually matters.

1. Understanding the product.

2. Understanding the customer.

3. Understanding the domain.

Thinking through the difficult problem that didn't disappear simply because AI made execution easier.

There is also an emerging reason to be careful about how much thinking we hand over. A controlled experiment with 130 participants found that high cognitive offloading to an AI assistant produced the best immediate decision accuracy and fastest decisions, but weaker subsequent skill development than the no-AI control. Moderate and lower offloading conditions supported stronger skill gains.[4]

I do not take this as an argument for using AI less. I take it as a reminder that immediate performance and accumulated human capability are not always the same thing.

The purpose of AI should not only be to help us do more.

It should also give us the capacity to understand the important things better.

I don't mean that human intelligence has literally become a commodity.

But access to general-purpose intelligence is clearly becoming cheaper and more widely available.

That changes the question.

When everyone can access powerful tools for writing, coding, analysis, research and ideation, simply having access to those capabilities stops being much of a differentiator.

Other things become more important.

- Attention.

- Context.

- Depth.

- Judgement.

The ability to stay with something long enough to understand what isn't obvious from the first answer.

AI can give you an answer in seconds.

It cannot give you years of accumulated context in seconds.

That is why I increasingly believe focused work will become more valuable, not less, as AI becomes more capable.

In an age where breadth is becoming extraordinarily cheap, depth may become one of the things that remains expensive and valuable.

[1] Zheyuan (Kevin) Cui, Mert Demirer, Sonia Jaffe, Leon Musolff, Sida Peng and Tobias Salz - “The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments with Software Developers.”

Randomized field experiments involving 4,867 developers at Microsoft, Accenture and an anonymous Fortune 100 company. The researchers report a 26.08% increase in completed tasks among developers with access to an AI coding assistant, with higher adoption and greater gains among less-experienced developers.
https://www.microsoft.com/en-us/research/publication/the-effects-of-generative-ai-on-high-skilled-work-evidence-from-three-field-experiments-with-software-developers/

[2] Sophie Leroy - “Why Is It So Hard to Do My Work? The Challenge of Attention Residue When Switching Between Work Tasks.”

Organizational Behavior and Human Decision Processes, 2009, 109(2), 168–181. Two experiments examined the difficulty of transitioning attention between tasks and found that attention can remain on unfinished prior work, affecting subsequent task performance.
https://www.sciencedirect.com/science/article/pii/S0749597809000399

[3] Prasanna B. Tambe - “Reskilling the Workforce for AI: Domain Expertise and Algorithmic Literacy.”

Management Science, published online September 2025; Volume 72, Issue 1 (January 2026), pp. 515–537. The study provides evidence that AI and algorithms complement domain expertise and can create greater value when domain experts themselves can interpret and apply algorithmic tools.
https://pubsonline.informs.org/doi/10.1287/mnsc.2022.03968

[4] Heng Ding, Yizhi Shen, Jing Chen and Ping Wang - “More vs. less cognitive offloading from AI assistants: impacts on novices’ collaborative performance and skill development.”

Information Processing & Management, available online; Volume 64, Issue 1, January 2027, Article 105046. In a controlled experiment with 130 participants, the high-offloading AI condition produced the highest immediate accuracy and fastest decisions but weaker subsequent skill development than the no-AI control.
https://www.sciencedirect.com/science/article/pii/S0306457326004371