In fact, their research shows that in some cases, pay has gone up. For instance, bookkeepers and accounting clerks saw computers take over many of their tasks between 1980 and 2018. Yet even as total employment in these roles fell by a third, their real hourly wages rose by nearly 40%.\u00a0<\/p>\n
The reason: If automation removes the simpler parts of a job, the work that remains often demands more expertise \u2014 which can make that work more valued and better paid because fewer people are qualified to do it.\u00a0<\/p>\n
But when automation targets the specialized tasks instead, the job may become easier for others to enter, increasing competition and putting downward pressure on wages.<\/p>\n
Like bookkeepers, inventory clerks were also highly exposed to computerization. But as automation targeted their most expert tasks, and employment in the role more than doubled, their real wages fell by 13%.<\/p>\n
\u201cJobs can either become more or less numerous, depending on what types of automation happen to the job,\u201d said Thompson, a principal research scientist at the\u00a0MIT Initiative on the Digital Economy<\/a>. \u201cIf the thing that gets automated is the easiest task, that leads to a narrowing of the pool of people that do it. But they are paid more.\u201d<\/p>\n The research shows that when business leaders think about how AI will change work, they should think beyond who is exposed to automation and also consider how it changes what people need to know to do a job.\u00a0<\/p>\n How automation alters roles and wages\u00a0<\/p>\n In seeking to understand why and how the nature of work has changed over four decades, the authors developed an \u201cexpertise framework\u201d to measure how much expertise different tasks require. Rather than relying on job titles or opinion surveys, they analyzed the language used to describe the work itself.<\/p>\n Task descriptions that include specialized terms \u2014 like those for operating diagnostic imaging equipment or forecasting economic trends \u2014 tend to require more expertise. Simpler tasks, such as boxing lightweight items or greeting customers, do not.<\/p>\n Using this language-based approach, the authors tracked how the average expertise level of more than 300 occupations changed between 1980 and 2018.\u00a0<\/p>\n They found that when simpler tasks disappeared, jobs became more specialized \u2014 and often better paid \u2014 even as employment declined. But when automation removed the more expert tasks, wages tended to fall as more people moved into the role.\u00a0<\/p>\n \u201cTaxi drivers, for example, once relied on deep knowledge of local streets, which was a real differentiator,\u201d Thompson said. \u201cBut with the arrival of GPS, that expertise was automated. The result is a more commoditized taxi service: lower wages, but many more drivers.\u201d\u00a0<\/p>\n This shift can create opportunities as well. \u201cThere will be new professions that open up to people,\u201d Thompson said, \u201cbecause automation removes the hardest parts that used to be out of reach.\u201d<\/p>\n <\/p>\n AI Executive Academy<\/p>\n In person at MIT Sloan<\/p>\n Register Now<\/p>\n <\/a><\/p>\n Practical implications for employers<\/p>\n The effects of automation require different management responses, depending on what kind of work is being automated, Thompson said.<\/p>\n If automation drives wages up but shrinks the number of people needed in a role, companies can often adapt gradually \u2014 for instance, by not replacing workers when they leave. \u201cThat\u2019s something you can manage through attrition,\u201d Thompson said.If wages go down and the job becomes less specialized, the adjustment is harder. Lower pay can hurt worker morale and make recruitment and retention more difficult.\u00a0<\/p>\n \u201cDepending on which tasks are automated, you\u2019ll see different effects in different parts of your workforce,\u201d Thompson said. \u201cYou want to think about that ahead of time because it calls for different responses from the company.\u201d<\/p>\n In some cases, leaders could design more specialized, high-value roles that retain expert tasks. In others, automation could be used to make certain jobs easier to access and expand the talent pool. Either way, companies have more control than they might assume.\u00a0<\/p>\n That variability is central to the study\u2019s argument. The authors do not suggest that automation is always good or bad for workers \u2014 only that its impact is more nuanced than many people assume.\u00a0<\/p>\n The research indicates that automation doesn\u2019t just reduce the need for human work. Sometimes it concentrates it around the tasks that are hardest to replace, and when it does, the value of that remaining human expertise can rise \u2014 and with it, wages, too.<\/p>\n This article is based on work by Neil Thompson and David Autor. Thompson is a principal research scientist at MIT Computer Science and Artificial Intelligence Laboratory and the MIT Sloan Initiative on the Digital Economy, and director of the MIT\u00a0FutureTech lab<\/a>. He studies technological innovation and firm strategy. Autor is a professor of economics at MIT and co-director of the\u00a0MIT Stone Center on Inequality and Shaping the Future of Work<\/a>. His research explores labor-market consequences of technological change.<\/p>\n","protected":false},"excerpt":{"rendered":"open share links close share links As automation accelerates and large language models continue to improve, the dominant…\n","protected":false},"author":2,"featured_media":36742,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[13],"tags":[72,61,60,150],"class_list":["post-36741","post","type-post","status-publish","format-standard","has-post-thumbnail","category-jobs","tag-business","tag-ie","tag-ireland","tag-jobs"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/36741","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/comments?post=36741"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/36741\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media\/36742"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media?parent=36741"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/categories?post=36741"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/tags?post=36741"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}![\"A](\"https:\/\/www.newsbeep.com\/ie\/wp-content\/uploads\/2025\/09\/ai-executive-academy_0.jpg\")
<\/p>\n![\"MIT](\"https:\/\/www.newsbeep.com\/ie\/wp-content\/uploads\/2025\/09\/MITSloanLogo_ExecutiveEducation_Horizontal_WHITE_web.png\")
<\/p>\n