Robots aren’t replacing everyone, but a quarter of U.S. jobs will be severely disrupted as artificial intelligence accelerates the automation of existing work, according to a new Brookings Institution report.

Thursday’s report from the Washington think tank says roughly 36 million Americans hold jobs with “high exposure” to automation — meaning at least 70 percent of their tasks could soon be performed by machines using current technology. Among those most likely to be affected are cooks, waiters and others in food services; short-haul truck drivers; and clerical office workers.

“That population is going to need to upskill, reskill or change jobs fast,” said Mark Muro, a senior fellow at Brookings and lead author of the report.

Tendencies in automation

The report cites “six basic tendencies in the workings of automation and its interplay with human labor that help in assessment.” They are:

Automation substitutes for labor

If a machine can do a task currently done by humans, it will do it with greater precision, speed, and at a lower cost—but there are limitations to such substitution.

Machines substitute for tasks, not jobs

A job is a collection of tasks, and even under the most aggressive scenarios, it is unlikely that machines will substitute for all tasks in any one occupation.

Automation also complements labor

Workplace activity that isn’t taken over by automation is complemented by it—making the remaining human tasks more valuable.

Automation can increase demand, creating jobs

Automation-driven cost and quality improvements can increase demand to a degree that offsets any would-be job losses.

Capital and labor augmentation spurs innovation

When machines handle routine, time-consuming activities, human capacity is freed-up to create new products and new tasks.

Tech possibility is not the same as tech reality

There are many reasons why technological adoption falls short of potential, so it is a mistake to equate technological potential with likely projected outcomes.

Source: Brookings

Read the full report online.

Muro said the timeline for the changes could be “a few years or it could be two decades.” But it’s likely that automation will happen more swiftly during the next economic downturn. Businesses are typically eager to implement cost-cutting technology as they lay off workers.

Though the United States is in the middle of its second longest expansion in history, and jobs data suggest that the economy remains healthy, many business leaders and economists have suggested in surveys that the United States could slip into a recession in 2020. In addition, the partial government shutdown has been creating anxieties about a downturn.

Some economic studies have found that similar shifts toward automating production happened in the early part of previous recessions — and may have contributed to the “jobless recovery” that followed the 2008 financial crisis.

But with new advances in artificial intelligence, it’s not just industrial and warehouse robots that will alter the American workforce. Self-checkout kiosks and computerized hotel concierges will do their part.

Most jobs will change somewhat as machines take over routine tasks, but a majority of U.S. workers will be able to adapt to that shift without being displaced.

Impact on wages, employment

In terms of determining the net impact of automation on employment and wages, MIT economist David Autor provides a simplified framework. In it, he highlights three primary dynamics:

What technology doesn’t replace, it complements

Workers who supply tasks that are substituted by machines are more likely to benefit from automation than are workers who supply tasks that machines can complete.

Wages will be determined by the ease with which roles in demand can be filled

Wage gains for the remaining tasks completed by humans will be larger when the barriers to entry (e.g. education, training, certification) are higher.

The number of jobs in an industry will be determined by the complex interaction of automation-driven price, quality, and wealth changes

Automation drives cost and quality improvements for products, as well as productivity and wealth increases for workers, which can mitigate some labor displacement.

Source: Brookings

Some chain restaurants have already shifted to self-ordering machines ; a handful have experimented with robot-assisted kitchens .

Google this year is piloting the use of its digital voice assistant at hotel lobbies to instantly interpret conversations across a few dozen languages. Autonomous vehicles could replace short-haul delivery drivers. Walmart and other retailers are preparing to open cashier-less stores powered by in-store sensors or cameras with facial recognition technology.

The changes will hit hardest in smaller cities, especially those in the heartland and the Rust Belt, according to the Brookings report. The risk is highest in Indiana and Kentucky, where some counties have nearly half the workforce employed in the labor-intensive manufacturing and transportation industries. The changes will also disproportionately affect the younger workers who dominate food services and other industries at highest risk for automation.

“Restaurants will be able to get along with significantly reduced workforces,” Muro said. “In the hotel industry, instead of five people manning a desk to greet people, there’s one and people basically serve themselves.”

Many economists find that automation has an overall positive effect on the labor market, said Matias Cortes, an assistant professor at York University in Toronto who was not involved with the Brookings report. It can create economic growth, reduce prices and increase demand while also creating new jobs that make up for those that disappear.

But Cortes said there’s no doubt there are “clear winners and losers.” In the recent past, those hardest hit were men with low levels of education who dominated manufacturing and other blue-collar jobs, and women with intermediate levels of education who dominated clerical and administrative positions.

In the future, the class of workers affected by automation could grow as machines become more intelligent. The Brookings report analyzed each occupation’s automation potential based on research by the McKinsey management consulting firm. Those jobs that remain largely unscathed will be those requiring not just advanced education, but also interpersonal skills and emotional intelligence.

“These high-paying jobs require a lot of creativity and problem-solving,” Cortes said. “That’s going to be difficult for new technologies to replace.”