Human progress has always advanced in step with technology, and every new stage has reshaped work. Step by step, it lightened the load on the individual — and still let prosperity rise. Less work, more prosperity: for centuries, the dependable rule.
Artificial intelligence could break that rule for the first time. Not by itself, but in concert with demographics. As the workforce shrinks and the algorithm takes over what people once did, prosperity could fall for the first time rather than rise. None of this is inevitable. It is a matter of design.
Design work deliberately, or it falls to the logic of short-term optimization — and that logic takes the path of least resistance: whatever is easiest to automate. But automation on its own only shrinks. It replaces jobs, and with them, purchasing power.
The alternative is not to automate, but to augment. The first question then is not what we can strip out, but what we can now do that we couldn’t before. What once lay out of reach.
Duolingo is a good example. With generative AI, every user now has what was once unaffordable: a personal tutor. Real conversations, scenarios to rehearse, an explanation the instant a mistake is made. No human could offer that to millions at this price. It is not simply a more efficient language course — it is a form of learning that wouldn’t exist without the technology. And that is precisely what made Duolingo grow.
And as long as we are redesigning work, we may as well think further.
Many people don’t fail at the task itself — they despair at the conditions around it. Mounting complexity stalls them: decisions take longer and longer, never shorter. Unclear ownership, missing goals, ever more meetings. It is dispiriting, and it costs productivity — not because people don’t try, but because the system holds them back.
This is where the greatest untapped reserve lies. AI need not only take over tasks; it can expose what blocks execution and help strip out the friction. Not more activity for its own sake — less resistance. To augment means this too: making organizations able to execute again, so that work takes effect again.
How do we get there?
First: augment, don’t automate. Don’t ask what the machine replaces; ask what it lets you achieve more of. Tie every automation decision to a single question: what will people do instead — and do more of? If there is no answer, it is shrinkage, not growth.
Second: shape the human part. Strengthen what the machine cannot replace: judgment, creativity, care, empathy. The things that make us singular. But take care — a working world made only of demanding tasks is no gift; it is relentless strain. To strengthen the human part is also to keep it bearable. And it means giving a say to those who do the work. Redefine work from above, and you forfeit the trust of the very people you rely on.
Third: reward human work anew. This begins with recognition. With being seen. People want to be acknowledged for their contribution, not vanish into an org chart. And they must be paid for it — fairly and transparently.
None of this gets easier — and that is where the outlook begins. The more that demanding work, too, passes to the machine, the more urgent the question becomes: what will we live on? As value comes increasingly from capital, data, and technology, and less and less from labor, wages alone will, at some point, no longer suffice. Redistribution through taxation will reach its limit as well. Then a stake in productive capital belongs on the table. Employee ownership decouples income from wages. Not a new idea — but one worth rediscovering, and carrying further.
Because the direction is clear. The question is not whether we sit in the office or at home. The question is whether we design work — or leave it to the machine.