AI changed the tools. What about the work?

A report that took a week now takes an afternoon. A first draft that took an afternoon now takes a minute. Code, summaries, translations, slide decks and research notes appear on request. For anyone who works at a keyboard, the last few years have felt like a sudden change in gravity.

And yet, look at the calendar of an average knowledge worker in 2026 and it looks much like the calendar of 2016. Standing meetings. Status updates. Approval chains. The same number of hours, the same expectations of availability, the same sense of running slightly behind.

The tools changed quickly. The work around them did not.

Saved time has to go somewhere

When a task becomes faster, the time it used to occupy does not disappear. It gets reassigned. In most organisations the default destination is more of the same task, or more tasks of a similar kind. If one person can now write three proposals in the time it took to write one, the natural managerial instinct is to ask for three proposals.

Economists have a name for a version of this. When an engine becomes more efficient, people often end up using more fuel, because cheaper energy invites more driving. Something similar happens with attention. Cheaper output invites more output. The inbox fills at the speed of generation.

Efficiency without a decision about what to do with it tends to become intensity.

This is not a failure of the technology. It is a failure of design. Nobody decided that the hours saved by AI would be spent on deeper thinking, rest, training, or time with customers. Nobody decided anything. So the existing system absorbed the gain in the only way it knew how.

Accelerating the wrong things

Consider the meeting. AI can now transcribe it, summarise it, extract action items and send the notes before anyone has left the room. Useful, certainly. But a cleaner record of an unnecessary meeting is still a record of an unnecessary meeting. The tool makes the ritual cheaper to perform, which may make it easier to keep.

The same pattern repeats elsewhere. Emails written by software are answered by software. Reports generated in seconds are skimmed by tools that summarise them in seconds. At some point it becomes reasonable to ask whether the report needed to exist.

Acceleration is powerful when the underlying process is sound. When it is not, acceleration mostly helps us arrive faster at outcomes nobody particularly wanted.

Neither revolution nor threat

Public conversation about AI and work tends to swing between two moods. One says everything is about to be transformed and anyone who hesitates will be left behind. The other says jobs are about to vanish and the only sensible response is fear. Both moods share a strange passivity. They treat the future of work as weather, something that happens to us.

A more useful stance is curiosity paired with scepticism. What does this tool actually do well? What does it do badly? What new problems does it create, such as errors that look plausible, or a flood of content that nobody has time to read? And who gets to decide how the benefits are distributed?

These are organisational questions more than technical ones. They concern management, incentives, trust and time. Answering them requires people who understand the work, not only people who understand the software.

Changing the structure, not just the speed

Imagine treating a major technological shift as an occasion to redesign, rather than an occasion to accelerate. A team might ask which of its recurring processes still serve a purpose now that information is easier to produce and retrieve. It might decide that some status meetings can disappear entirely. It might shorten the week for certain roles, or protect long blocks of uninterrupted time, or spend saved hours on the parts of the job that machines handle poorly: judgement, relationships, care, taste.

None of these choices is automatic. Each involves trade-offs and some risk. But each represents a deliberate decision about what the work is for, rather than an unexamined assumption that faster is always the goal.

If technology gives us new capabilities, why are we using them to preserve old ways of working?

The question is not rhetorical. There are real reasons organisations hold on to familiar structures. They are legible. They make coordination predictable. Changing them requires effort and a willingness to be wrong for a while.

Still, the moment when tools change is one of the rare moments when habits are loose enough to reconsider. Every previous wave of technology offered that opening, and most of the time we used it to do the same things faster. The opportunity now is to do something less common: decide, on purpose, what the work should become.

The software will keep improving either way. Whether the work improves with it is up to us.