AI promised to make us productive. Instead, it made us busy. And there's a growing pile of evidence that most of us can't tell the difference.

The feeling versus the reality

Last year, the nonprofit METR ran a randomized controlled trial with experienced open-source developers. The result was striking: when developers used AI tools, they completed tasks 19% slower than without them. But here's the twist, when asked how AI had affected their speed, those same developers estimated it made them 20% faster.

That's a nearly 40-percentage-point gap between perception and reality. They weren't lying. They genuinely believed AI was helping. The tools felt productive. The autocomplete was snappy. The suggestions kept flowing. But none of that translated into actually getting things done faster.

This is the busy idiot in its purest form: someone who is doing more, seeing more, generating more, and yet accomplishing less.

The productivity paradox, in numbers

The developer study is far from an isolated case. Across industries, a pattern keeps emerging: AI increases output without increasing outcomes.

A 2025 PricewaterhouseCoopers survey of 4,454 CEOs across 95 countries found that only 12% of companies said AI had grown their revenues and reduced their costs. Meanwhile, 56% said they were getting "nothing out of it."

A study from the AI consulting firm Section surveyed 5,000 white-collar workers. Two-thirds of rank-and-file employees said AI saved them zero to two hours per week. But more than 40% of executives claimed AI was saving them over eight hours a week.

The Upwork Research Institute's 2024 survey painted an even starker picture: 96% of executives expected productivity gains from AI, while 77% of employees said it had actually increased their workload. 39% reported spending more time reviewing AI-generated content. And 71% reported burnout.

The numbers don't add up, unless you accept that what leaders are measuring isn't productivity. It's activity.

Workslop and the illusion of output

There's a term for the AI-generated material that masquerades as useful work: workslop. CNBC coined it to describe AI content that looks productive but lacks the substance to actually move a task forward.

Workslop is everywhere. It's the AI-drafted email that technically answers the question but misses the point. It's the slide deck that's beautifully formatted and strategically hollow. It's the proposal that has the right numbers but the wrong assumptions.

And the plurality of it, about 40%, comes from peers. But at least 16% comes from above, from managers who use AI to generate plans and documents that their teams then spend hours fixing.

A Workday study confirmed the downstream effect: much of the time employees saved by using AI tools was offset by extended reviews of AI-generated content. The time didn't disappear. It just moved to someone else's calendar.

When speed becomes drag

Hamilton Mann, writing for IMD, described a case that captures this dynamic perfectly. A global B2B software company rolled out an AI productivity transformation for 2,800 go-to-market employees. On paper, everything worked. Marketing could draft campaigns instantly. Sales reps could generate proposals in minutes. The dashboards glowed.

But within six weeks, email volume to prospects had tripled. Unsubscribes climbed. Response rates dipped. Sales reps were spending more time skimming AI drafts than crafting relevant messages. Legal and brand teams added new review steps. Average time-to-send increased. Downstream teams drowned in "fast" documents built on wrong assumptions.

The executive dashboards celebrated a 6% productivity gain. The actual win rate was flat or declining.

This is what happens when you optimize a single node in an interdependent system. Local speed creates systemic drag. The bottleneck doesn't vanish, it migrates.

The real cost: skill erosion

The subtler danger is what happens to the people in the loop. When AI handles the drafting, the summarizing, the first pass at analysis, humans start to lose the muscles those tasks once built.

Research from the Center for Economic Policy Research found that AI users showed the lowest brain activity during writing tasks and increasingly relied on copy-pasted content. The group writing without AI showed the highest brain connectivity and creativity-related neural activity.

Mann observed the same pattern in organizational settings: managers were coaching less and curating more, triaging machine-made outputs to find the ones that mattered. Core skills like probing, reframing, and cross-team sense-making atrophied quietly.

And there's an insidious framing effect. If the AI summarizes a sales call around price objections, the team discussion gravitates there, even when the real issue is trust or fit. The tool sets the frame. The humans follow.

Why executives don't see it

There's a structural reason the perception gap skews the way it does. Executives benefit from AI's output without bearing the cost of its cleanup. They generate more plans, more strategies, more communications. Their calendars feel lighter. Their dashboards look better.

But as Molly Kinder of the Brookings Institution put it, workers face a different calculus entirely. The upside of using AI is murky, while the downside risks are clear and often existential. Why get better at a tool that might replace you?

This asymmetry in motivation explains a lot. Executives are incentivized to believe AI is working. Workers are incentivized to be skeptical. And neither group has a clean view of the actual impact on business outcomes.

How to stop being the busy idiot

The fix isn't to abandon AI. It's to stop confusing motion with progress. A few principles help:

Measure outcomes, not output. The number of emails sent, documents drafted, or tasks completed tells you nothing about whether the business is moving forward. Track win rates, cycle times, customer satisfaction, and decision quality instead.

Audit the downstream cost. Every hour AI "saves" upstream might create two hours of review, correction, or confusion downstream. Map the full workflow before celebrating efficiency gains.

Protect the thinking work. Not every task should be delegated to AI. The tasks that build judgment, the ambiguous ones, the ones that require synthesis across contexts, are often the ones most worth doing yourself.

Be honest about the vibes. Feeling productive is not the same as being productive. If you can't point to a concrete business outcome that improved because of AI, the gains might be illusory.

Design for the system, not the node. AI transformations fail when they optimize individual tasks without considering how those tasks connect. Speed in one place means nothing if the rest of the organization can't absorb it.

The uncomfortable truth

AI is an extraordinary technology. It will reshape how we work, and in some domains it already has. But right now, for most knowledge workers, the primary effect of AI is not productivity. It's the feeling of productivity, a pleasant hum of activity that papers over the absence of real progress.

The busy idiot isn't stupid. The busy idiot is seduced by the same thing that seduces all of us: the comfort of doing something, anything, instead of sitting with the harder question of whether it matters.

The best use of AI might not be doing more. It might be doing less, but doing it well.

References