AI Promised Focused Work. It Delivered More Email.

For years, the sales pitch for AI has sounded almost identical.

Less admin. Less grind. More time for deep work.

More room for better thinking.

More space for strategy, creativity, and judgement.

It's a compelling promise.

It's also not what the data is showing.

The most useful thing about the latest ActivTrak analysis, reported in the Wall Street Journal, is that it cuts straight through the hype. It looked at 164,000 workers across 443 million hours of digital work activity, spanning 1,111 employers. The result was not subtle.

There was not a single activity category where AI actually saved users time. As the report put it, the data is unambiguous: AI does not reduce workloads.

That should force a much more honest conversation.

Because if AI was meant to create more room for focused work, what many organisations have actually created is more digital noise.

We didn't remove work. We sped it up

This is the part we keep coming back to.

Most organisations didn't redesign work when AI arrived. They kept the same meetings, the same approval layers, the same reporting loops, the same threads, the same back-and-forth. Then they added tools that can generate more drafts, more summaries, more updates, and more content for someone else to check.

So the machine speeds up activity.

But it doesn't automatically reduce the amount of work flowing through the system.

That is exactly what the ActivTrak data reflects. It's not showing teams suddenly freed up for deep thinking. It's showing teams drowning in more communication, more messages, and more fragmented attention. Focused work is shrinking while coordination work expands.

AI didn't remove friction. In many places, it multiplied it.

The inbox got louder, and focused work kept falling

One of the most striking details in the ActivTrak analysis is that this is not just a one-off dip in concentration. It's part of a longer trend. The length of the average focused, uninterrupted work session fell by 9%, and focused work hours dropped by another 2%. The report describes this as the continuation of a three-year downward trend, with the share of time spent in the zone falling to 60% in 2025.

That matters because focused work is not a nice extra. It's where strategic thinking, judgement, and complex problem-solving actually happen. If that time keeps eroding, then even productivity gains on paper can mask a much bigger performance problem underneath.

You can produce more.

You can reply faster.

You can push out more drafts.

But if people have less uninterrupted time to think, the quality of decisions starts to suffer, even while the volume of activity keeps climbing.

That's not leverage.

It's a hidden tax.

The cost is not just workload. It's burnout and cognitive strain

The Wall Street Journal piece also points to a University of California, Berkeley study that helps explain why this matters. It found that as employees tap into AI-driven efficiency gains, they often end up taking on more work and using the time that would have been natural breaks to do more prompting and more task switching. The warning is simple: people still need time to recharge, or they become less productive, not more.

That fits neatly with the Harvard Business Review research on what it calls "brain fry" too. The issue is not AI in the abstract. The issue is the kind of work pattern AI is creating. When people are spending their day overseeing tools, toggling between outputs, checking machine-generated work, and managing more inputs than before, the pressure does not disappear. It shifts into cognitive load.

So yes, AI can absolutely help with repetitive work.

But if the broader system then uses those gains to pile on more tasks, more outputs, and more communication, people do not feel lighter. They feel saturated.

The data is not anti-AI. It is anti-lazy implementation

This is where we think too many leaders get defensive.

The point is not that AI never helps.

The point is that most organisations are introducing it in ways that make work denser instead of better.

Even in the Harvard Business Review research, there's an important nuance. When AI is used to substantially reduce time spent on routine or repetitive tasks, burnout scores are 15% lower. That tells us the problem is not the tools alone. The problem is what we do around them. When AI genuinely removes toil, it can help. When it adds oversight, more workload, and more tool-juggling, it drives strain instead.

That's why the best-performing organisations are not just using AI more.

They're doing the opposite of what most firms are doing.

They're redesigning the work.

The organisations getting results are simplifying, not layering

The organisations getting real leverage from AI are doing the opposite of what most are doing: smaller teams, clearer ownership, fewer handoffs, and faster decision loops. They're not layering AI onto the same cluttered operating model. They are redesigning how decisions get made and how work moves.

That's the distinction that matters.

A cluttered system with AI is still a cluttered system. In fact, it is often a faster cluttered system.

If you want focused work back, you don't get there by adding another assistant, another agent, or another dashboard on top of an already noisy environment. You get there by asking harder questions.

What work should disappear entirely?

What decisions need fewer handoffs?

What approvals are no longer necessary?

What communication is just status theatre?

What part of the process is genuinely improved by AI, and what part is simply producing more outputs for people to react to?

Those are design questions.

And most organisations still are not asking them early enough.

Why this matters more than the next tool purchase

A lot of AI investment still assumes the next gain will come from the next tool.

But if the underlying pattern is more email, more messaging, and less focused work, then adding another tool to the same system is often just adding another source of coordination load.

The Harvard Business Review research makes that risk even clearer. Productivity rises as people move from one tool to two, but the gains shrink with a third, and after three, productivity starts to dip. Multitasking does not become smart just because AI is involved.

That is why this conversation needs to shift from tool acquisition to work design.

Not, what else should we buy?

But, how should work actually function now?

Where do humans need to stay in the loop?

Where are we creating hidden rework?

Where has AI sped up output without improving decisions?

Where are we increasing activity while reducing the space people need to think?

Those questions are not secondary to AI strategy.

They are the strategy.

Before you buy another AI tool, redesign how your teams actually work

The data is clear enough now that leaders should stop pretending this is a temporary wobble.

AI did not magically hand people more focused time.

In many organisations, it handed them more email.

More messages.

More checking.

More cognitive switching.

And less room to think.

That does not mean the answer is to pull back from AI.

It means the answer is to get more serious about work design.

The organisations seeing real gains are not the ones chasing every new tool.

They're the ones rethinking workflows, reducing handoffs, clarifying ownership, and protecting the conditions that make good work possible in the first place.

So before you buy another AI tool, redesign how your teams actually work.

And if you want to know where AI is already creating hidden friction in your organisation, that is what the AI Readiness Experiment is built to find. It puts people on a real work problem for ten days and captures what actually happens: where AI creates value, where it creates rework and checking burden, and where the conditions around the work are the thing that needs to change.