The attention problem isn’t solved. And AI is already here.
There have been times when I preferred to spend my break scrolling through social media rather than having a coffee with my colleagues.
Work stopped for a few minutes. My attention stayed on a screen.
I think back to those moments when I talk about digital habits. They touch on the space we make for other people, for breaks and for what matters to us in a working day.
This question runs through my book Be Present: how can we regain a sense of choice in our digital habits? It remains open. We are still trying to make room for concentration, conversations without screens and evenings that do not turn into more scrolling.
Now we also need to learn to work with tools that can write, analyse and make suggestions for us.
In my first article, I explained why time saved with AI does not automatically become time available. The next part of that question concerns our attention: what conditions do we need to understand, check and decide?
We start with habits that are already well established
At work, a single message can be enough to pull us away from a task. But sometimes we open our inbox without a notification, simply because we are struggling with a sentence or waiting for a reply.
My OFF/ON surveys continue in 2026. In the dataset already consolidated, covering 2,138 respondents, 73% say they find it difficult to stay focused.[4]
This figure describes the people surveyed. It does not allow us to attribute that difficulty to any particular technology. It does, however, remind us that concentration already needs our attention as AI use grows.
Gloria Mark, a researcher at the University of California, Irvine, has studied these shifts in attention for years. She points out that we also interrupt ourselves. Her observations lead her to call for changes in individual practices and in how work is organised.[1]
This matters. Someone who knows how to turn off notifications can still work in a team where an immediate reply is expected.
Training in a new tool takes place within that daily reality. Old habits continue while new ones take hold.
AI brings another difficulty
Take a simple situation. You ask an assistant to prepare a recommendation. It gives you three options, the advantages of each and a convincing conclusion.
You have something to get started with sooner. You still need to determine whether you asked the right question, whether the information is accurate and whether the recommendation fits your situation.
That means knowing the subject, holding the relevant points in mind and following an argument long enough to see its limits.
The document is ready. Your attention may be elsewhere.
This is where the two issues meet. We rely on human judgement to check AI outputs, while working conditions may already make that judgement difficult to exercise.
Writing “human review” at the end of a process leaves open the question of whether the time and attention to do it are available.
The research calls for a closer look at how we use AI
A study published at CHI in 2025 surveyed 319 professionals about their use of generative AI. Greater confidence in the tool was associated with less self-reported critical thinking. Participants also described a shift in effort towards checking and integrating the answers.[2]
This survey does not show that AI causes people to lose intellectual abilities. It invites us to examine what we still do ourselves when we use it.
A BCG survey published in June 2026 offers a complementary perspective. Among the regular AI users surveyed, 67% say it improves their job satisfaction. Within that same group of regular users, 41% report increased cognitive load.[3]
These are self-reported perceptions. They show the value of tracking several dimensions at once: enjoyment of work, the gains achieved and the effort required.
I use AI myself to research, produce and experiment. It can help me understand a subject, examine another hypothesis or work through a difficulty. I also want to be able to explain the choices I make.
Working practices need to evolve alongside the tools
We therefore have two things to work on in parallel: the habits that scatter our attention and the way we entrust certain tasks to AI.
A team can learn to write better prompts while continuing to treat everything as urgent. An employee can produce an analysis quickly and struggle to defend its conclusions in a meeting.
For me, helping people adopt these tools means looking at what they can actually do with the result: explain it, discuss it, correct it and use it appropriately.
This concerns working conditions as much as skills. Asking for a thorough review means making room for it in the day.
A starting point: three lines before the first prompt
For a task that calls for analysis or a decision, I suggest a small experiment: formulate your own starting point before asking AI.
Take two minutes to write:
- The problem I am trying to solve.
- My view or hypothesis at this stage.
- What could change my mind.
Then use AI to explore, challenge or build on that initial thinking.
Your hypothesis may be wrong. The tool may help you correct it. Those three lines simply give you a position to test, rather than an answer to accept because it sounds convincing.
It is a practice to try, adapt to the task and discuss as a team.
Attention remains a work in progress. AI already requires us to think about the role we want to retain at work: as people who can understand what they do and take responsibility for it.
For human, with technology.
What about your organisation?
When your teams use AI, do they have the conditions they need to examine its answers and decide what to do with them?
The Digital Load Check-up helps make the points of friction around digital tools and AI visible, then identify three priorities for action, remotely, in three weeks.
Frequently asked questions
Why does attention still matter when we use AI?
Because understanding a result, checking its sources and judging its relevance still require attention. Automating a task does not guarantee that these conditions will be met.
Does AI reduce our critical thinking?
The study by Lee and colleagues identifies an association between confidence in AI and less self-reported critical thinking. It does not demonstrate a general loss of our abilities or a causal effect.[2]
How can we maintain our own thinking when using AI?
For an analytical task, we can start by writing down the problem, our hypothesis and what might challenge it, then use AI to examine that thinking. This is the practice proposed in this article.
Gabriel Pitt is a speaker, author of Be Present and founder of OFF/ON®. He advises organisations on digital work, attention and AI.
Sources and method
[1] UC Irvine, interview with Gloria Mark, “Regaining Focus in a World of Digital Distractions”, 26 January 2023. An introduction to her research and her book Attention Span; the observations described do not constitute a universal measure of human attention capacity.
[2] Hao-Ping Lee et al., “The Impact of Generative AI on Critical Thinking”, CHI 2025. A survey of 319 professionals, with 936 examples of AI use. Self-reported data; the observed associations do not establish a causal effect.
[3] BCG, “AI Is Reshaping Jobs Faster Than Companies Are Reshaping Work”, 3 June 2026, a summary of the AI at Work 2026 survey. Total sample: 11,749 workers across 14 markets. The figures of 67% and 41% refer to regular AI users. This is a consultancy survey based on respondents’ own reports.
[4] OFF/ON surveys, conducted since 2024 and continuing in 2026. The figure cited, 73%, relates to an already consolidated dataset of 2,138 respondents. Figure published on Gabriel Pitt’s website. Self-reported data describing those surveyed; this result does not demonstrate a causal effect of digital technology or AI on concentration.
The AI-assisted recommendation scenario is an illustration. The three-line exercise is a practical suggestion from the author; it has not been evaluated as an intervention in the studies cited.