AI is getting very good at removing friction from work. It can shorten research, speed up analysis, generate first drafts and reduce the time it takes to get from a question to an answer.

That’s the opportunity.

The harder question is what happens when the friction being removed was also helping someone think, learn, test an assumption or improve the work.

Because not all friction is waste. Sometimes the friction is the part that forces us to think more carefully before we decide.

As AI makes more of the work easier, the challenge is no longer simply how much friction we can remove. It’s knowing which friction was getting in the way, and which friction was making the work better.

Not all friction is the same

There’s friction that adds effort without adding value. Broken processes, unnecessary approvals and repetitive work belong in that category. That’s friction worth removing.

But there’s another kind. The back-and-forth that happens when someone challenges an idea. The review that catches something the first pass missed. The first attempt that gives someone something to learn from.

That kind of friction may slow the work down, but it can also improve it.

So before we remove a step, we need to understand what was happening inside it.

Speed isn’t the same as being right

A friend recently told me about a team using AI as part of a security testing exercise. The tool surfaced a number of potential vulnerabilities. Some were legitimate. Others were errors.

It took people with the right experience to review the findings and determine which ones actually represented a risk.

The AI made part of the work faster. Human judgment made the output reliable. They needed both.

That distinction matters.

AI can reduce the effort it takes to produce an answer. But getting to an answer and knowing whether it’s the right answer are two different things. The second still depends on experience, context and knowing when something doesn’t quite make sense.

And that raises a more important question: how do people build that judgment in the first place?

How judgment gets built

A lot of professional judgment is built by working through problems, getting challenged and learning from people with more experience.

Over time, you start to recognize patterns and know when something doesn’t quite add up. That experience becomes judgment.

AI can help someone who already has that judgment move faster. But for someone still building it, we need to think more carefully about what happens when AI starts doing more of the first draft, research, analysis or problem solving.

This matters even more when the thinking behind the work is becoming less visible. In many workplaces, people already spend less time watching experienced colleagues work through problems. Add AI into the mix, and we may increasingly see the finished answer without seeing the reasoning that produced it.

The answer isn’t to preserve unnecessary manual work. It’s to be more intentional about how people build experience and learn from one another when more of the process happens out of sight.

The capacity AI gives back

This is where the opportunity gets bigger than efficiency.

If AI turns three hours of work into thirty minutes, we haven’t just saved time. We’ve created capacity.

What happens to that capacity is a choice. We can use it to produce more, or we can reinvest some of it into making the work better.

That might mean spending more time challenging the answer rather than producing another one.

That’s the dividend AI can create. Not just more output, but more room for the parts of work that benefit from judgment, collaboration and deeper thinking.

Collaboration is friction too

Some of the best work I’ve seen during transformation didn’t happen because everything moved quickly. It happened because people were brought into the conversation early.

They questioned assumptions and explained how a change would actually land in their part of the organization. They surfaced things that weren’t obvious in the project plan.

That takes longer than making the decision in a small room and communicating it later. But that’s the point.

AI can help people arrive at those conversations better prepared. It can help them organize information, explore options and get to a point of view faster.

But the conversation still matters because that’s often where the work gets challenged and improved.

What was the friction doing?

This may be the more useful question as we decide where AI fits into work:

What was the friction doing?

If it was creating delay without adding value, remove it.

But if that friction was helping people build judgment or improve the work, understand what you’re removing before you automate it away. 

That doesn’t mean the old process needs to stay. It means we need to decide where that value will happen instead.

Better use of the AI dividend

The goal isn’t to preserve work simply because that’s how it’s always been done.

Make the easy things easier. Remove the friction that adds no value. Automate the steps that don’t need human attention.

But understand what you’re removing along with them.

If AI gives us back time, the opportunity isn’t simply to fill that time with more work.

It’s to use some of it to make the work better.