This spring, Meta told thousands of U.S. employees that software on their work computers would begin capturing mouse movements, clicks, and keystrokes. Meta needed this data, they said, to help train its AI agents on real examples of how people work.
And it’s not just Meta that needs this kind of data. Any company trying to automate workflows with artificial intelligence needs access to how humans actually perform these tasks. McKinsey, for instance, created its internal generative AI, Lilli, by drawing on the expertise that had been developed in the firm over its long history.
The director of design for Lilli made the underlying logic explicit. For most of McKinsey’s history, she said, “our knowledge was with our experts.” Lilli is the attempt to take that knowledge out of the experts and make it an organization-wide resource.
From the perspective of the organization, this makes perfect sense. But the benefit is not quite so obvious from the perspective of the expert. In May, a month after news of Meta’s click-tracking broke, the company laid off 8,000 people and gave AI as the reason. More broadly, more than 175,000 people have been laid off in tech in 2026, and the main reason given for these layoffs has been AI’s rapidly improving capabilities.
So let us put it in simple terms. Companies want their employees to give them their expertise to train AI agents. And once that expertise has been adequately transferred, companies will get rid of their people and let AI agents take over.
This is not like asking turkeys to vote for Christmas. This is like asking turkeys to buy the presents, decorate the tree, cook the meal, and put the drunk uncle safely to bed. It is rational for the turkeys to resist. Business leaders need to approach their own goal differently, and doing so begins with understanding how knowledge transfer works.
Knowledge transfer needs cooperation
In a recent article for Harvard Data Science Review, my coauthors and I argued that AI has helped solve the technical challenge of knowledge retrieval. Instead of having to manually look for needles in very disorganized and intimidatingly large haystacks, we can now use large language models to search unstructured data and answer questions in ordinary language.
This is an enormous improvement. But there is a very important limitation that no amount of AI can overcome: Better retrieval only helps with knowledge that made it into the system in the first place. And when it comes to expertise, a lot of knowledge never reaches the system at all.
Now, some expertise is easy to capture—any expertise that has been codified, that has been described, that has been captured in words or in processes is relatively simple to access. So things like templates, playbooks, and workflows can be turned into data straightforwardly and used to train AI agents. Artifacts like these can be taken and incorporated without the expert’s cooperation.
But there are parts of expertise that cannot be harvested purely from the outside. For example, you can record that an experienced salesperson abandoned a lead after the third call, but not that a particular hesitation in the client’s answer reminded her of three deals that later collapsed. You can observe a senior engineer departing from the standard troubleshooting procedure, but not that she recognized a combination of weak signals that, from experience, usually points to a deeper problem. Nor does the record show the alternatives the expert considered and rejected, the exceptions she has learned to recognize, or what she would have done had one detail been different.
Those things can become training data, but only if the expert helps make them visible: by explaining her reasoning, comparing cases, identifying exceptions, and articulating the intuitions experience has taught her. That means the most valuable targets for knowledge transfer cannot simply be extracted. They depend on the expert’s choosing to cooperate.
The limits of asking
Well, then, why not ask your experts? You can, and you should. But it is no panacea. Well before the rise of AI, organizational psychologists analyzed a phenomenon they called “knowledge hiding”; research suggests that roughly half of employees intentionally withhold or disguise what they know from their own colleagues.
Knowledge hiding is prevalent in the context of AI too. In one recent survey of 4,000 workers, 35% said they are hoarding knowledge specifically because they fear being replaced by AI, and 38% are reluctant to train colleagues in areas they consider personal strengths. Relatedly, a three-wave study of 348 knowledge workers working alongside AI found that employee-AI collaboration raises fears of job insecurity, and job insecurity drives knowledge hiding.
Companies cannot coerce their way to a solution to this problem, because compliance can be performative. This is what the research calls “evasive hiding.” Evasive hiding is cooperation-shaped: the employee attends every session and answers every question, but they nevertheless withholds the information that really matters.
Observation is not enough
If asking fails, watching looks like the natural next step, which is precisely the logic behind Meta’s keystroke capture program. But this does not offer a complete solution either.
History gives us an eerie parallel. A century ago, Frederick Taylor brought stopwatches onto machine-shop floors as part of an effort to turn workers’ practical know-how into something management could observe, measure, and standardize. The machinists understood what was at stake. They coordinated around a deliberately restricted pace, discouraged one another from revealing how quickly the work could really be done, and thereby denied management an accurate picture of their productive capacity.
Taylor called this kind of behavior “systematic soldiering” and condemned it as one of industry’s great evils even while acknowledging the reasons behind this behavior: Workers understood that once management discovered that a job could be done faster, this would lead to higher demands on their labor without equivalent increases in pay. The workers’ reaction was a rational attempt to protect the value of their knowledge and effort. The ensuing battles over time studies helped fuel years of labor conflict.
In sum, surveillance is not a clean source of data because the interests of the watcher and those being watched are not aligned. Instead, it changes the work being performed, and can also lead to conflict. And finally, the part of expertise that leaves no outer traces is in any case invisible to surveillance methods, and no amount of watching a worker will surface it.
Helping your experts help you
There is a world of difference between cooperation and compliance. And the fact that true knowledge transfer requires cooperation changes the organizational problem. Rather than asking, “How do we capture more?” organizations need to ask, “Under what conditions will our experts want us to succeed?”
Here are four things business leaders can do to create the conditions that encourage experts to cooperate:
1. Be explicit about the bargain. Tell employees what information is being collected and why. If the plan is to reduce head count, say so; if the plan is to move people into different work, explain what that means in practice. People are much more likely to withhold when they suspect the real purpose of the exercise is being concealed, because the uncertainty provides motivations to conceal a broader range of knowledge.
2. Address job insecurity directly. If employees reasonably think that successful knowledge transfer will lead to their losing their jobs, no amount of “trust building” will fix the incentive problem. Companies need to make credible commitments about what happens to the people whose expertise is transferred.
3. Keep experts attached to the system. Make them stewards who audit, correct, and update the workflows. This also helps with insecurity because experts can see a future role for themselves rather than a one-time knowledge extraction event.
4. Give them a stake in the upside. If employees are being asked to transfer expertise that makes the company less dependent on them, give them a tangible benefit for doing so. That might mean bonuses for documenting and validating expertise or creative solutions like profit-sharing when the system produces measurable savings. If the company captures the upside of the transfer while the employee absorbs most of the risk, there is little rational reason to cooperate.
Success requires partnership
Your experts are not passive sources of training data; they are participants who understand what knowledge transfer may mean for them. And if you want their best judgment, you cannot merely extract it—you have to make cooperation rational. That requires transparency, security, an ongoing role, and a share in the upside.
The companies that treat their experts as a resource to be exploited rather than as partners to collaborate with will collect plenty of data. But they will fail to collect the deep expertise they actually need.
The most important skills in today’s workplace are judgment and discretion.
This is because the cost of output has never been cheaper. With models like Fable producing yet another step-function increase in capabilities, it’s easier than…
When I was in college at the University of South Carolina, I thought I wanted to work in media, but I didn’t really know what that meant or how someone like me was supposed to get there.
I’m from a small town in South Carolina. I didn’t have family…
In 2019, I published a book whose English title would be something like From Graft to Craft. I argued that craftsmanship isn’t a sector but a model of work that could in principle apply to any job. The distinction I built the book around was between graft (fragmented, deskilled, industrialized toil) and craft (a finished piece of work you can put your name to). To work as a craftsperson means autonomy, responsibility, creativity, and mastery of a result you can be proud of. A developer, a nurse or a consultant should all be able to work that way. Instead, too many of us are handed the deskilled version and feel alienated or bored.
A fascination that never quite matched reality
For years, the story of graduate-degree holders walking away from bullshit jobs to become bakers, cheesemakers, brewers, restaurateurs kept coming back in the media. These reinventions fascinate everyone, journalists first of all, forever hunting for seductive stories about rediscovered meaning, hands deep in the grease. But our collective fascination always said more than the phenomenon warranted. The neoartisans were, statistically, a rounding error, a marginal trend inflated into a movement because it flattered a widely shared dream as well as nostalgia for a wholeness of work that probably never existed in the first place.
We’ve been here before
All of this was already in the air more than 15 years ago. Matthew Crawford’s Shop Class as Soulcraft (2009) made the philosophical case. The former think tank director who quit to fix motorcycles argued that manual work demands real intelligence and offers a psychic wholeness that knowledge work, for all its prestige, often doesn’t. The cultural craze was real. It even reached software, where “software craftsmanship” became a movement with a manifesto, a way for developers to insist their work was a craft and not an assembly line.
The craze was everywhere but in employment statistics. The trades actually kept struggling to recruit. The bullshit jobs kept multiplying. Software craftsmanship stayed the conviction of a passionate minority while most engineering work industrialized anyway. The much-mythologized flight of elite graduates into the trades was, in the end, statistically negligible.
Trades in high demand. White collars in the doldrums.
Start with the trades. Skilled-trade shortages have become structural across the developed world: aging workforces retiring faster than young workers can replace them, and recruitment rated difficult across the board. Roofers, welders, electricians, auto body workers and butchers are hard to find. In my small country alone, France, at least 150,000 craft jobs sit unfilled, and roughly 300,000 artisanal businesses will need new owners over the next decade as large generations retire. The U.S. is staring at the same wall: More than one in five construction workers is over 55, and the construction industry alone needs an estimated 349,000 net new workers this year, most of them just to replace retirees. Across all skilled trades, one industry analysis projects a shortfall reaching 2.1 million unfilled positions. All of this arrives just as climate change multiplies the need for construction, energy retrofits, insulation, and the physical adaptation of cities.
Meanwhile, the future white-collar workers can’t get hired and are increasingly pessimistic about their future. In the U.S., recent college graduates—those between 22 and 27—are now more likely to be unemployed than the workforce as a whole, something the New York Fed’s data, going back to 1990, had never recorded before. Their jobless rate sat at roughly 5.6% through mid-2026, well above the national rate. Gallup’s global polling makes the mood vivid: The U.S. now has the widest gap of any country between how young and old adults view the job market, with just 43% of Americans between 15 and 34 calling it a good time to find work, 21 points below their over-55 counterparts. The steepest pessimism of all sits with the most educated young Americans still looking for full-time work, the people a college degree was supposed to protect.
AI, both the cause and the alibi
Artificial intelligence is probably part of the reversal, and certainly part of the alibi. When a company slows hiring, “we’re automating” is a more flattering story than “we overhired in 2021 and can’t afford the downturn,” and the two are genuinely hard to separate. The evidence that AI is doing real damage is mounting: Stanford’s Erik Brynjolfsson and coauthors found that since late 2022, early-career workers between 22 and 25 in the most AI-exposed roles—software developers, marketing staff—saw a 16% relative decline in employment, even as older workers in the same jobs held steady. But the picture is contested. Anders Humlum, a University of Chicago economist, points to payroll data showing that the companies spending the most on AI actually expanded their entry-level headcount.
What isn’t contested is where the damage concentrates: the bottom rung. The tasks that language models handle best (summarizing documents, drafting first-pass code, producing the standard memo) are those once handed to juniors so they could learn the trade. Now that these tasks can be automated, the short-term move is to stop hiring the person who used to do them. The trouble is that a ladder without its bottom rungs is one you can’t climb onto. Companies are trimming a real cost today while dismantling the system that produces their future senior talent.
Gen Z is already doing the math
The cost of college makes the calculation even more brutal in the U.S. The New York Times has documented a Gen Z turning to trade schools to shield themselves from both student debt and AI. Enrollment in public vocational schools reportedly rose nearly 20% between 2020 and 2025. AI can write an email, but it can’t install a sink or weld a valve, they reason. The main obstacle they name is other people’s judgment. The vocational path is not seen as prestigious. Only a tiny minority of parents want their child to choose a manual trade.
But the young increasingly see the trades as meaningful, as offering secure prospects and fostering independence. Many say they’re ready to move toward technical or hands-on work. For that trend to have a future, we need to support apprenticeship, make it into a machine that inserts young people. Alas, apprenticeship hasn’t grown lately . . .
Why this time might actually be different
The craftsmanship craze of the 2010s stayed marginal because the numbers didn’t back it up. Prestige and pay kept flowing to the knowledge jobs. Today, however, three forces could actually make craftsmanship into the model of the future.
The first is AI. As it industrializes white-collar work and floods it with slop, generic, deskilled, interchangeable output produced at scale, the comparative value of work that can’t be faked and slopped goes up. You need a weld to hold and a pipe to not leak. Craftsmanship could become an economic choice in the late 2020s.
The second is physical demand. It’s enormous. The housing crisis gripping much of the developed world is, among other things, a construction problem. We are not building fast enough, and much of what we have built needs work. Climate change compounds it. The energy transition runs through the built environment: Insulation, retrofits, heat pumps, and the wholesale renovation of aging housing stock require a lot of work. Adaptation to a hotter, more volatile climate means reinforcing, rebuilding, and redesigning cities and infrastructure for conditions they were not made for. None of this can be prompted into existence.
The third is demographic. It will force us to widen the definition of craft. As populations age, we won’t only need roofers and electricians but also a lot of people in care work. We don’t usually file caring professions under “craftsmanship” but we should. They demand exactly what defines craft: embodied skill, judgment, a result that depends on the person doing it, and a physical presence. If craft is work that accumulates in the body and can’t be faked, then bathing a frail patient without hurting them, or reading distress in a person, is craft of a high order. Recognizing it as such may be the most consequential revaluation of all.
Nearly half of Gen Z workers, 48% to be exact, frequently or consistently feel they are not enough. That statistic should stop every leader who reads it, because it isn’t a motivation problem, a work ethic problem, or a talent problem. It’…
By the time Katie returned to work after parental leave, her daughter had started attending a day care close to her apartment. But on days when the day care was closed or she needed childcare coverage, there was a childcare center conveniently located…
Do artificial intelligence tools such as ChatGPT eliminate jobs, create new ones, or both? There’s a lot of speculation regarding this question, but little solid data in the U.S. to answer it.
I’m a sociologist who researches AI, among other topics…
When it comes to commuting or remote work, half of U.S. workers prefer working from home. In fact, 40% of respondents in a Harvard Business School survey said they would accept a pay cut of 5% or more for the ability to telecommute. It’s also an incre…
Majoring in computer science used to be considered a safe career bet, ensuring stable pay and work. After the 2008 recession, many students turned to the field’s promising outlook: Between 2008 and 2024, the number of computer science degrees grew by …
For the last year, the Justice Department has been quietly probing high-profile companies over their diversity, equity, and inclusion policies. Deloitte settled one such investigation earlier this week, agreeing to pay $21.5 million to address allegat…
