Expertise, Not Just Task Automation, Will Shape the Future of Work

Rather than asking, “How many jobs will disappear?” the sharper question is, “Does automation lift or lower the expertise bar in this field?” That shift in expertise – not raw job counts – is what determines who benefits and who gets squeezed, and by how much.

The story we’ve been presented on automation is a simple one: machines replace repetitive jobs, people move to the new ones, wages and jobs regulate the situation. It's a tidy model – perhaps too tidy. According to a recent study by economists David Autor and Neil Thompson from MIT Blueprint Labs, what matters is not the amount of tasks automated, but the level of expertise they require. And when you change the focus off tasks and onto expertise, the labor market no longer appears as a continuum, but as a tug-of-war with unlikely winners and losers.

The Puzzle That Wouldn't Go Away

Take accounting clerks and inventory clerks. Computerization struck both jobs. Ideally, they both ought to have been punished equally: reduced daily operations, reduced manpower, reduced pay. But that's not what happened. In the long run, accounting clerks experienced increases in their wages despite a decline in employment numbers whereas inventory clerks experienced decreases in wages despite an increase in their ranks.

Such a breakdown could not be fully attributed to the old paradigms of task displacement and automation exposure. Something more incisive is suggested by Autor and Thompson: an expertise framework. The insight is almost too simple: when the automation of a task eliminates it, that either raises or lowers the expertise required on what remains. And wages and quantity of work change with that shift in the expertise demanded – not merely with the amount of work.

Once inexpert jobs are automatized, the occupation, by definition, becomes more expert-intensive. Consider bookkeeping. All lower-order data entry and reconciliation moved over to computers, leaving behind openings in more sophisticated judgment, decision-making and analysis. That increased the mean level of expertise needed.

Expertise Changes – Jobs Changes

What occurs in the labor market? The remaining tasks become rarer and require more specialized skills, which increases wages. However, the number of workers who can meet that increased requirement decreases. Jobs are lost, although wage checks get chubby.

That is why accounting clerks turned into a smaller yet more highly paid group. The demand of their labor became narrow, the price of those who were able to clear the bar became high.

The reverse can also happen. In inventory management, automation swept away the more complicated elements – such as reorder point computation or pricing policy – and left the comparatively unskilled physical operations of counting or stocking. The expertise bar dropped.

Because of this, there were additional workers who would undertake the rest of the work. Jobs were created but salaries were lowered. Markets now opened up to a larger labor pool thereby eliminating scarcity value. The paradox of this: automation not only created more jobs, but also de-rewarded them.

Quality not Quantity

Much of the automation discussion has been preoccupied with the number of tasks machines are performing. Autor and Thompson demonstrate that, in fact, the type of task is far more important. It is one thing to lose ten routine and low-expertise tasks, but another to lose ten complex and high-expertise tasks. The first one leaves behind a harder job but with more pay but less seats; the second one leaves behind an easier job with more seats but less pay.

The expertise lens resolves riddles that have stumped economists for years: why automation of routine jobs sometimes raised wages in occupations where it hollowed out jobs, and why in other cases it had decreased wages despite increasing the number of jobs. Experience becomes the secret lever, shifting the results in the opposite directions.

There is, however, a bleak implication in this change of attitude.

First, automation does not universally constitute skills bias or deskilling; it may indeed constitute both at once, depending on the tasks that automation is applied to in the job. Second, wage differentials will tend to increase not only between sectors but between jobs, as certain jobs will become closed to more highly skilled workers and others will be opened to low-skill workers.

There's also a cultural angle. With the expertise threshold, the prestige of an occupation may increase or decrease. Bookkeeping transformed to financial analysis with increased remuneration and status. inventory work, deprived of its more challenging components, drifted towards commoditization. Automation does not just redefine paychecks, it reshapes identities.

Policymakers and businesses need to frame strategy on this. Education and training become important levers in case automation increases the level of expertise. Should it reduce them, the struggle becomes the need to provide good wages and movement in an increasingly packed labor market.

The moral here is obvious: the future of work is not going to be more or less based on how many tasks the machines do, but rather the skills that they will leave behind. And that slight difference may be the difference between automation being a ladder or a trap.

  Autor, David, and Neil Thompson. 2025. Expertise. Discussion Paper #2025.07. Cambridge, MA: Massachusetts Institute of Technology, Department of Economics, and National Bureau of Economic Research.