Lately, many people have been saying the same thing.

AI feels slower.

Reasoning models pause longer. Thinking modes take more time. The spinner lingers.

Most people treat this as a performance problem.

Something is wrong with the tool. The model is overloaded. The system needs to be faster.

But research presented at CHI, the Conference on Human Factors in Computing Systems, suggests something more interesting is happening.

Slower AI is not just revealing how models think. It is revealing how humans behave when thinking takes time.

The paradox leaders should pay attention to

The CHI study examined how people co-created ideas with AI using two different interaction modes.

• Breadth-first AI produced faster responses, more options, and easier scanning.
• Depth-first AI produced slower responses, fewer outputs, and deeper reasoning.

Participants consistently felt that breadth-first AI was more creative. It felt energising, expansive, and productive. Momentum was visible.

But when researchers evaluated the outputs themselves, the results told a different story.

The more creative, novel, and surprising ideas came from depth-first AI.

Speed improved perception.

Depth improved results.

This is not just an AI insight. It is a work insight.

The real variable was not the model

One of the most overlooked findings in the research was this.

The difference in outcomes did not come from the AI alone. It came from what people did while the AI was thinking.

Some participants treated the pause as dead time. Others used it as thinking time.

Those who stayed engaged during the delay reflected on the problem, refined their intent, and prepared better follow-up direction. They were not waiting passively. They were working.

Participants who used the waiting period actively:

• Explored ideas more deeply

• Gave clearer follow-up prompts

• Reported a stronger sense of control over the outcome

Waiting was not a gap in the process. It was part of the process.

Passive Waiting vs Active Waiting

This introduces a distinction that matters at work.

Passive Waiting looks like checking email, switching tabs, or filling the gap with noise. The pause becomes something to escape.

Active Waiting looks like reviewing the question, examining assumptions, and preparing the next move. The pause becomes productive.

The research showed that people who engaged in Active Waiting produced better outcomes and felt more ownership over the work.

The AI did not simply think harder. The human did too.

W.A.I.T. – A practical habit for AI thinking time

To make this usable in day-to-day work, a simple rule helps when AI slows down. Not as patience. As intent.

W: Work the question
Re-read what you asked. Clarify what a good answer would actually look like.

A: Audit assumptions
Notice what you may be taking for granted or deciding too quickly.

I: Inventory inputs
Identify missing context, constraints, or examples that could sharpen the response.

T: Thread the next move
Prepare the follow-up prompt before the answer arrives.

W.A.I.T. turns AI latency into thinking leverage. Instead of fighting the pause, you use it.

Control is another hidden paradox

Another insight from the research is uncomfortable for many leaders.

When AI responds quickly and offers many options, people feel more in control. When AI slows down and explores deeply, people feel less control, even though the results are often better.

This mirrors leadership dynamics. Tight control feels safe. Letting go creates space for emergence.

With AI, as with people, creativity often improves when leaders loosen their grip and allow thinking to unfold.

This is not a tool problem. It is a mentality problem.

In the 7 Drivers of AI Effectiveness, this behaviour sits under one driver: Mentality.

Mentality describes when and how naturally people bring AI into real work, not whether they know how to prompt.

Low Mentality shows up as using AI only at the end, treating it as a last-minute fixer, or abandoning it when it slows down.

High Mentality shows up as using AI throughout the task, staying engaged during reasoning, and knowing when to prioritise speed and when to allow depth.

This is where speed and quality stop being opposites and become a sequencing decision rather than a trade-off.

Why this matters at the team level

At a team level, these behaviours compound.

Two teams can use the same AI tools, with the same access and training, and still produce very different outcomes.

The difference is not better prompts. It is how they behave during pauses. How they treat delays. How they balance velocity with judgment.

This is exactly what the Team AI Effectiveness Scorecard is designed to surface. It gives leaders an observation-based view of how teams perform across seven drivers, including Mentality, alongside Velocity, Quality, Continuity, and Scalability.

The goal is not to audit tools. It is to reveal behaviour patterns.

A final reframe for leaders

As AI systems reason more deeply, pauses will become more visible.

The question is no longer why AI is slower.

The better question is what your team does when thinking takes time.

Because in that moment, AI is not the only thing being tested.

Your mentality is.

If you want a clearer picture of how your team is actually working with AI, not just whether they are using it, the Team AI Effectiveness Scorecard provides a practical starting point. It helps leaders spot where small behaviour shifts can unlock much larger outcomes.

Same tools. Same models. Very different results.