AI: time saved does not become time available

“A wealth of information creates a poverty of attention.” — Herbert A. Simon, 1971

In Be Present, I suggested rewriting the old “commute, work, sleep” as “scroll, work, scroll, insomnia”. Four years later, we need another version: “prompt, review, meeting”.

Am I exaggerating a little? Certainly. But not that much.

AI has delivered on part of its promise: it produces things quickly. A report in three minutes, ten versions of an email, a summary of a forty-page file. Yet many of the managers I meet tell me the same thing: they produce more, without having more time to think.

The paradox is only apparent.

The bottleneck has moved

Writing, calculating, formatting: these tasks used to take up a large part of office work. AI now speeds up many of them.

But another limit remains: our ability to read, compare, judge and decide. Our working memory can hold only a few items at a time. It has not received an upgrade.

Every text intended for use needs to be reviewed. Each version may need to be compared. Every important answer needs to be checked before it circulates. Producing more also means creating more material to absorb.

We are no longer short of hands. We are short of headspace.

Herbert Simon put it into words in 1971, long before the first smartphone: when information becomes abundant, attention becomes scarce. AI did not invent this problem. It can accelerate it.

What the studies show

In a 2024 Upwork survey, 77% of employees who use AI say it has added to their workload. 39% spend more time reviewing or correcting generated content.

40% of office workers surveyed in the United States say they received “workslop” during the previous month: AI-generated content that looks polished but is insufficient to move the work forward. Respondents estimate that resolving each case takes around two hours on average (BetterUp Labs and Stanford Social Media Lab, 2025).

For eight months, researchers at the University of California, Berkeley observed a technology company of around 200 people. Work did not become lighter there; it intensified: more tasks, several activities running in parallel, and work spilling into breaks and evenings (Ranganathan and Ye, 2026).

Let us be clear: AI does save time. But a tool can speed up a task while adding work around it: verification, coordination, new requests.

The real question is: how much of the gain remains once all that work is counted?

What I see in practice

My own surveys show that digital workload goes beyond AI alone. Of 184 managers and employees surveyed in 2025, 82% feel it slows their progress on their priorities.

During a recent Check-up, 11 out of 18 respondents described an expectation of quick replies, even when nothing was urgent. Only 3 out of 17 felt that focus was valued as much as availability. Small numbers, but they put words to habits that teams often take for granted.

A manager recently told me a story that captures the problem well. Before a training session, he had asked an employee to define their objectives. A few lines would have been enough. He received four pages. Well written, well structured… and almost unusable: workslop in all its glory.

Rather than give a long speech, he responded with the same weapon: four AI-generated pages. The message got through.

Eight pages for a request that could have been answered in a few lines.

What interests me in this story is the lack of instructions. One simple rule, “for your objectives, a few lines written by you”, would have avoided two rounds of reading and a misunderstanding.

Théo Compernolle compares multitasking to a Swiss Army knife with every blade open. With AI, we have added a few more blades. And made them faster.

One way to keep the time you save

Agree on the rules as a team.

This is often where individual gains are lost. Remember those four pages. When is AI welcome, and when do we expect a few lines written by the person themselves? Who checks generated content before it circulates? What format is enough to make a decision? What response time is actually expected? You can start small: one team meeting, three common uses, a few explicit rules. For example: check the facts before forwarding a text, put the main request at the beginning of a message, specify which situations require a quick reply. Without shared rules, everyone saves time on their own task… and costs others time.

This is not about being for or against AI. It is about choosing what we do with the time it frees up.

Pro human. Pro technology.

What about your organisation?

If your teams produce more but make less progress on what matters, look at the often unwritten rules around their tools: what needs checking, what is worth circulating, what is urgent and what can wait.

The Digital Load Check-up makes these points of friction visible and helps you choose three priorities for action, remotely, in three weeks. Discover the Check-up →

Frequently asked questions

Does AI really reduce workload?

It reduces the time spent on some tasks, such as writing, summarising or formatting. But this gain does not guarantee a reduction in overall workload: verification, coordination and new tasks must also be counted. In the 2024 Upwork survey, 77% of surveyed employees who use AI say it has added to their workload.

Why can AI feel tiring?

Because it shifts part of the effort towards reading, comparing and validating, and encourages more iterations and tasks running in parallel. It draws on your judgement, which itself depends on your attention: it is hard to judge well what you cannot focus on.

What can an organisation do to retain the time savings from AI?

Define shared rules: which uses are useful, who checks generated content, what level of quality is enough, what response times are realistic. Then check whether these gains actually serve the priorities. The Digital Load Check-up helps identify the three priorities to address first.

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

Herbert A. Simon, “Designing Organizations for an Information-Rich World”, 1971.

Upwork Research Institute, From Burnout to Balance: AI-Enhanced Work Models for the Future, 2024. Self-report survey in the United States, United Kingdom, Australia and Canada involving 2,500 people (625 employees, 625 freelancers, 1,250 executives). The 77% figure refers to employees who use AI.

BetterUp Labs and Stanford Social Media Lab, Workslop: The Hidden Cost of AI-Generated Busywork, 2025. Self-report survey of 1,150 full-time office workers in the United States.

Aruna Ranganathan and Xingqi Maggie Ye, “AI Doesn't Reduce Work—It Intensifies It”, Harvard Business Review, February 2026. Eight months of qualitative research in a single company; the findings do not apply to all organisations.

OFF/ON surveys, 184 managers and employees, 2025. Self-reported data, non-representative sample. The Check-up mentioned is a separate engagement; the number of responses varies by question.

Théo Compernolle, Swiss Army knife comparison cited in Gabriel Pitt, Be Present, 2022.