The Productivity Paradox: We Automated Everything and Feel Busier Than Ever

The productivity paradox is the quiet contradiction that runs underneath modern working life: we built machines that do our work for us, and somehow we ended up busier. The tools are astonishing. A phone in my pocket drafts documents, translates languages, books travel, and answers questions that would once have taken a trip to a library. Whole categories of drudgery have been automated away. And yet the promised payoff — the shorter day, the lighter week, the leisure our grandparents were told the machines would buy — has stubbornly refused to arrive. If anything, the opposite happened. The inbox is fuller. The pings are constant. The bar keeps rising. That gap between what the technology can do and what our lives actually feel like is the puzzle I want to sit with.

Two paradoxes wearing the same name

It helps to separate two things that get tangled together, because both are real and they are not identical.

The first is the economists’ version. In 1987 Robert Solow made the famous quip that you could see the computer age everywhere except in the productivity statistics. Businesses were pouring money into information technology, but the measured productivity gains — output per hour, across the whole economy — were disappointingly flat for long stretches. That is the classic productivity paradox: a mismatch between enormous investment in new technology and the sluggish numbers that were supposed to reflect it. Economists have argued about the causes ever since. Maybe the gains were real but mismeasured, because so much of what computers give us is hard to price. Maybe there is a long lag while organisations learn to actually use a general-purpose technology. Maybe the benefits were concentrated in a few industries and never spread. I hold these explanations loosely, because honest people who study this for a living still disagree, and I would rather hedge than hand you a tidy number I cannot stand behind.

The second version is the one you feel in your body at 9pm when a message arrives and you answer it. Call it the lived paradox. Regardless of what the aggregate statistics say, the personal experience of automation has not been more free time. It has been more expected output, delivered faster, with the expectation of near-constant availability layered on top. Both paradoxes matter. But it is the second one that keeps people up at night, and it is the one I think we can actually do something about.

What the tools actually gave us

Consider email. It collapsed the cost of sending a message to roughly zero and shrank the delay from days to seconds. That is a genuine, enormous efficiency gain. But watch what happened to the gain. Because sending became free, the volume exploded, and the burden shifted from the sender to the receiver — the person who now has to read, sort, and answer a hundred messages a day. The efficiency did not evaporate. It was captured and redistributed. Someone got the time back; it just was not the person staring at the screen.

Slack and its cousins did something similar to the working day itself. They dissolved the boundary between “at work” and “not at work.” The green dot that says you are available became a soft obligation to be available. Automation was supposed to reduce the amount of work a human had to do. Instead, in a lot of jobs, it raised the amount of responsiveness a human was expected to supply. The machine got faster, so the person was asked to keep pace.

The efficiency did not evaporate. It was captured and redistributed. Someone got the time back — it just was not the person staring at the screen.

This is the pattern I keep noticing across every wave of new technology, and I have written before about how technology gets captured: a tool arrives promising to lighten a load, and within a few years the load is heavier, only differently shaped. The saved effort does not sit there as a gift. It gets reabsorbed almost immediately, because someone with more power than the worker gets to decide what the saved time is for.

Where the saved time went

So where did all that dividend go? Not into thin air. I can trace at least four channels, and you have probably lived all of them.

Higher expectations. When a report that used to take a week can be produced in an afternoon, the deadline becomes the afternoon. The tool does not buy you the extra days; it resets the baseline of what “a reasonable amount of work” means. Standards ratchet upward to consume whatever slack the technology created, and they very rarely ratchet back down.

New tasks. Automating one job tends to spawn three others around it. Spreadsheets did not eliminate financial analysis; they let one person model twenty scenarios where before there was time for one, and then twenty scenarios became the expectation. The tool creates capacity, and the capacity gets filled with work that did not exist before.

Availability. The single biggest thing many of us have handed back is the boundary around our attention. The technology that lets you work from anywhere quietly becomes the technology that lets work reach you anywhere. That is a real cost, even though it never shows up on a payslip.

Rising standards. The finished thing is expected to look more polished, be more thorough, arrive more quickly, and never contain the kinds of errors the old tools made unavoidable. The bar for “good enough” climbs to sit right on top of whatever the newest tool can reach.

Notice that none of these are laws of physics. They are choices — mostly made by someone other than the person doing the work. And that is the heart of it. The productivity dividend is real. The question of who gets to spend it is a question about power, not about technology.

The capture lens

Here is the through-line I keep returning to. When a new tool makes work more efficient, that efficiency becomes a dividend that has to go somewhere. By default, it does not go to the worker as free time. It flows to the owner as profit, or to the customer as lower prices, or to the manager as higher output targets — unless the worker has the power to claim a share of it as time. Absent that claim, the dividend defaults upward. That is not a conspiracy; it is just what happens when the person who captures the gain is not the person who does the labour.

You can see this clearly if you ask a simple question of any productivity improvement: who takes it, who pays for it, and who could fight back? When a factory automates, the owner takes the savings, the displaced worker pays, and whether anyone fights back depends on whether workers have organised enough to bargain. The technology is the same in every case. The distribution of its benefits is a political outcome, decided by who holds leverage.

This is exactly why the great automation of the last few decades did not deliver the leisure it promised. The economist John Maynard Keynes famously predicted in 1930 that his grandchildren would work fifteen-hour weeks, because productivity would grow so much that the economic problem would essentially be solved. He was right about the productivity — output per worker did grow enormously. He was wrong about the fifteen-hour week, not because the dividend failed to materialise, but because almost none of it was taken as time. It was taken as more stuff and more profit, because the people deciding how to spend the dividend were not the people who would have enjoyed the extra hours.

Why more output always wins the default

There is a structural reason the dividend defaults to output rather than time. A firm that turns a productivity gain into more production can sell more and grow. A firm that turns the same gain into shorter hours for the same pay has, on its own balance sheet, simply raised its labour costs per unit. In a competitive market, the pressure runs relentlessly toward output. So without a deliberate countervailing force — a union, a law, a norm, a boss who actually decides otherwise — the leisure option quietly never gets chosen.

This is why the fight over working time keeps coming back, and why I find the experiments around the four-day week so interesting. They are, at bottom, an attempt to take a productivity dividend deliberately as time rather than letting it default into output. Where those trials have been run seriously, the striking finding is that output often holds steady while hours fall — which tells you the dividend was there all along, sitting inside the working week, waiting for someone to claim it. The four-day week is not a productivity trick. It is a decision about who gets to spend the gains.

The same logic sits underneath the return to office fights. When commuting and co-location stopped being technically necessary for a lot of knowledge work, a genuine dividend appeared — hours of commute, and control over one’s own space and time. The tussle over returning to the office is, in large part, a tussle over who gets to keep that dividend: the worker who saved the commute, or the organisation that would prefer the older arrangement. Once again, the technology settled nothing. Power did.

The AI turn — same paradox, higher stakes

Which brings us to the wave breaking over us right now. Artificial intelligence is being sold with the exact same promise every previous tool was sold with: it will do the tedious parts, and you will be freed for higher things. And I want to be clear that the capability is real. These tools genuinely can compress hours of work into minutes. The efficiency is not hype.

But if the pattern holds, the efficiency will do what efficiency has always done. It will become a dividend, and it will default upward, and most of us will find ourselves expected to produce more, faster, while being available for even more of the day — unless we do something different this time. I have written at length about whether these tools are net job creators or destroyers in will AI create jobs, and the honest answer is that it depends far less on the technology than on the choices made around it. The machine does not decide whether it frees you or drives you harder. People do.

The machine does not decide whether it frees you or drives you harder. People do. The productivity paradox is not a fact of nature — it is a fact about who gets to spend the gains.

The hopeful turn

Here is where I land, and it is genuinely hopeful, because the paradox is not a law of nature. If the reason automation never delivered leisure is that the dividend defaulted to output and profit, then the remedy is not to slow down the technology. It is to change who decides how the dividend gets spent — and to consciously choose to take some of it as time.

That is a choice available at several levels. As a society, we can make it through policy and through the slow renegotiation of norms about what a full working week even is. As organisations, leaders can decide — and some are deciding — that when a tool doubles a team’s effective output, part of that gain is returned as shorter hours or better pay rather than every last drop being converted into more work. As individuals, we have less leverage, but not none: we can defend the boundaries around our attention, refuse the premise that faster tools mean permanent availability, and be honest about which tasks the tool actually eliminated versus which ones we simply invented to fill the space it opened.

The productivity paradox has always been, at its core, a distribution question dressed up as a technology question. We did automate an enormous amount, and we do feel busier than ever, and both of those things are true at once precisely because the time we saved was never ours to keep by default. The heartening part is that “by default” is not the same as “by necessity.” With this next wave, we get another chance to answer the old question deliberately: when the machine hands us back an hour, do we let it be swallowed by higher expectations, or do we claim it as an hour? For the first time in a long while, enough people are asking the question out loud that the answer might actually be up for grabs. Whether we get more time, in the end, depends on whether we decide to take it.