Hyper Productivity but at What Cost

AI promised time savings. Automate the repetitive stuff, move faster on the meaningful stuff, and reclaim hours in the day.

Instead, an Upwork study found that 77% of employees say AI has increased their workload, not decreased it. Meanwhile, 96% of executives still expect AI to boost productivity.

There is a name for this pattern. The Jevons Paradox. It was first observed in 19th century England: when steam engines made coal more efficient to use, coal consumption did not fall. It rose. People found more things to do with cheap energy.

AI is doing the same thing to work. Making tasks faster does not mean there is less to do. It means we find more to do.

The productivity gains are measurable. So are the costs.

The Speed Is Real

The gains are genuine. A Harvard/BCG study of 758 consultants found that those using AI finished 12.2% more tasks, completed them 25.1% faster, and produced results that were 40% higher quality.

Small teams are shipping what large teams used to. Cursor hit $500M ARR with roughly 50 people. Bolt reached $20M ARR with 15 people in two months. Solo founders are building and selling companies in months that would have taken years with a full team.

BCG's 2025 survey found employees report saving about five hours per week with AI. That is a meaningful amount of time.

The question is: where is that time going?

New Capabilities, New Reach

One of the most notable effects of AI is how it expands what a single person can do.

Engineers are doing product and design work. Companies like Linear, Vercel, and Cursor are hiring "design engineers" who ship real interfaces, not just mockups. Tools like v0 let a developer upload a Figma screenshot and get working, interactive code back in minutes.

Product managers are writing code. Replit CEO Amjad Masad called PMs "some of the best vibe coders" because they know how to break problems into clear steps. CEOs are showing up to meetings with working prototypes they built themselves.

The "full stack employee" is emerging. LinkedIn replaced their Associate Product Manager program with an Associate Product Builder program, combining coding, design, and PM skills into one role. Their CPO Tomer Cohen put it simply: the old model of separate handoffs between PM, design, and engineering "is broken."

Shopify's CEO made AI usage a baseline expectation. Managers must prove why AI cannot do a job before requesting new headcount.

With AI tools, a single individual can punch above their weight, operate across disciplines, and produce output that once required multiple roles.

Now here's the caveat: When you can do everything, you start trying to do everything. And that is where the cost begins.

People Are Doing More

An 8-month UC Berkeley study embedded researchers with 200 employees at a U.S. tech company. They observed work patterns across engineering, product, design, research, and operations. Their finding was blunt:

"AI tools didn't reduce work, they consistently intensified it."

This aligns with what I have observed in my own use of AI tools. Three patterns stand out in how AI quietly increases workload.

Reviewing AI Work

Every AI output needs a human to check it.

This sounds minor until you realize how much review work adds up. A METR study found that developers felt 20% faster when using AI, but when actually measured, they were 19% slower. The gap? Time spent reviewing, correcting, and validating AI output.

BetterUp Labs and Stanford found that 41% of workers have encountered "workslop": AI output that looks polished but lacks substance. Each instance costs nearly 2 hours of rework. For a 10,000 person organization, that adds up to a $9 million annual productivity hit.

The work shifts from creating to reviewing, verifying, and correcting what the AI produces.

The "One More Prompt" Trap

AI removes the friction that used to give us natural stopping points.

Before AI, starting a task required effort. You had to open files, set up context, think through the approach. That friction was also a boundary. It gave you a reason to stop, take a break, go home.

Now? It takes five seconds to type a prompt. So you fire one off before lunch. One more before you leave. One more before bed. Each one feels small. But they add up.

The UC Berkeley researchers observed exactly this. Workers squeezed prompts into lunch breaks, meetings, and evenings. Conversations with AI felt more like chatting than working, which made it easier to blur the line.

"As they became accustomed to typing prompts during their breaks, their breaks no longer brought the same sense of recovery they once did."

The workday loses its natural pauses. Recovery time disappears. And you do not even notice because each individual moment feels effortless.

Running Multiple Jobs at Once

With AI, you can run several tasks in parallel. Write code while an AI generates an alternative version. Kick off one agent for research while another handles a refactor. Revive a long deferred task because the AI can "handle it" in the background.

It feels like leverage. In practice, it is compression.

You are not offloading cognitive work. You are packing more of it into the same hours. The American Psychological Association found that task switching costs up to 40% of productive time.

And there is a subtler cost. When AI handles the simpler, routine tasks, you are left with only the complex decisions. Your brain loses the "cognitive palate cleansers" that lower intensity work used to provide. You end up running at peak cognitive intensity all day with no breaks built in.

The Expectation Shift

Nobody formally raises expectations. But informal norms shift fast.

When your colleague uses AI to take on extra responsibilities, standing still feels like falling behind. Within months, doing what AI made possible becomes what is expected.

As one engineer in the UC Berkeley study put it:

"You had thought that maybe, oh, because you could be more productive with AI, then you save some time, you can work less. But then really, you don't work less. You just work the same amount or even more."

The Burnout Equation

These patterns have measurable consequences.

Upwork's 2025 research found a striking pattern: workers who report the highest productivity gains from AI are also the most burned out. 88% of them experience burnout. They are twice as likely to consider quitting.

The people getting the most out of AI are also the ones closest to the exit.

By month six of the UC Berkeley study, reports of burnout, anxiety, and decision paralysis had spiked among participants. Not because the work was bad. Because there was simply too much of it, with too few pauses and too little recovery.

Gartner's 2026 Future of Work Trends now lists "AI's biggest hidden cost: your employees' mental fitness" as a top trend for CHROs.

What Can We Do About It

The UC Berkeley researchers proposed a framework they call "AI Practice": deliberate habits that counteract the natural pull toward intensification. Combined with practical experience, here is what helps.

Build Intentional Pauses

Take conscious, intentional breaks. Your brain needs time to reset and recover, not just switch between tasks. Without that rest, quality drops and burnout builds.

Breaks also boost creativity. Research shows that new ideas form when the mind is not under load, not while staring at a screen.

And here is the bigger opportunity: use the time AI saves you to do what you always wanted to do. Pick up a new hobby. Learn something outside your field. Read widely. Those experiences feed back into your work in unexpected ways. The best solutions to hard problems often come from outside the problem itself.

Don't Let Speed Replace Judgment

Do not let AI speed dictate the pace of your judgment. Before finalizing a major decision, pause. Add one counterargument. Link back to the actual goal. Speed is not a substitute for thinking.

Protect Human Connection

Check ins. Shared reflection. Conversations with actual people about the work you are doing. Do not let AI become your only "colleague." The UC Berkeley researchers specifically warned against employees drifting into isolation with their AI tools.

Set Boundaries With the Tool

Time box your AI sessions. Close the laptop. The prompt can wait until morning. If you find yourself firing off "just one more" before bed, that is a signal, not a habit to lean into.

Measure Outcomes, Not Output

Volume is easy to inflate with AI. Lines of code, documents drafted, tasks completed. These metrics become meaningless when AI can produce them at scale. Shift to measuring impact: what changed, what improved, what moved the needle.

Closing Thoughts

AI is a powerful tool. The speed is real, and so are the new abilities it gives us. People are now building things that would have been out of reach just a year ago.

In my own case, I've used the time AI saves me to explore ideas and start projects that I would have once put off because they felt too time-consuming. That's where the real value lies. As I wrote in Lessons from Toyota: Looms to Cars, the biggest opportunity with AI is not doing the same work faster. It is using the shift to explore entirely new value streams and opportunities.

But the Jevons Paradox is playing out in real time. Efficiency is not reducing the workload. It is expanding it. A tool that increases output while reducing recovery time may not be improving productivity in a sustainable way.

The researchers at UC Berkeley said it best:

"The organizations that thrive will not be those that adopt AI most aggressively, but those that adopt it most thoughtfully."

The same goes for individuals. Use AI to do better work, not just more of it.