AI Could Reduce Labor Demand Equivalent to 36 Million US Jobs: 10 Most Exposed Fields

Artificial intelligence may transform the US labor market on a historic scale over the next decade.

But one of the biggest numbers in the debate requires careful interpretation.

McKinsey Global Institute estimates that AI-enabled automation and other technologies could reduce labor demand by the equivalent of roughly 36 million full-time jobs by 2035.

That is not the same as forecasting 36 million layoffs.

The firm's September 29 report argues that most of the disruption could be absorbed through job growth, changes within existing occupations and the creation of entirely new roles.

What the 36 Million Figure Actually Measures

McKinsey estimates that automation technologies could absorb about 54% of current US work hours by 2035.

Yet jobs are bundles of tasks.

Automating one task does not necessarily eliminate the employee who performs it.

Organizations may redesign jobs, increase output, move employees toward higher-value activities or retain workers for responsibilities that still require judgment, coordination and accountability.

After accounting for those effects, McKinsey estimates that labor demand could fall by the equivalent of about 21% of current work hours, or roughly 36 million full-time jobs.

The figure is therefore an economic estimate of reduced labor demand—not a head count of people certain to become unemployed.

Most of the Impact Could Be Absorbed Within Existing Occupations

McKinsey estimates that roughly 25 million of the 36 million job equivalents affected by automation could be offset by growth within the same occupations.

Workers could stay in their profession while their responsibilities change.

A financial analyst, software developer or administrative employee may continue to hold a similar title but spend much less time performing tasks that AI can automate.

The result may be job transformation rather than job elimination.

About 11 Million Workers May Need an Entirely New Occupation

The more disruptive outcome applies to the remaining group.

In McKinsey's base case, about 11 million US workers, roughly 7% of the current labor force, may need to move into a different occupation by 2035.

The estimate ranges from approximately six million to 16 million depending on the speed of automation adoption and its effect on labor demand.

Those workers would not simply need a new employer.

Many would need a substantially different type of work.

The Economy Could Still Create More Jobs Than Automation Removes

The same report projects that broader economic growth, demographic change and the expansion of the AI economy could create demand for more than 40 million jobs over the next decade.

McKinsey's base model puts that figure at roughly 41 million.

This creates a labor-market paradox.

There could be more jobs in 2035 than there are today, while millions of workers simultaneously struggle to find a practical route from shrinking occupations into growing ones.

That is why McKinsey frames the challenge primarily as one of workforce mobility.

Ten Occupational Groups With High Automation Exposure

The detailed estimates identify several occupational groups where a large share of work hours could be affected.

The public McKinsey summary rounds some adoption figures differently—for example, describing office support as roughly 80% automated and technology and analytics at 70%—so one-point differences can appear across presentations of the underlying data.

Office Work Faces Some of the Highest Exposure

Administrative jobs combine large employment numbers with highly structured tasks.

Scheduling, document processing, basic reporting, data entry and routine communication are increasingly within the capabilities of AI-enabled software.

That produces high technical exposure.

But even here, automating most activities does not necessarily mean eliminating most employees.

Remaining responsibilities can shift toward coordination, exceptions, customer interaction and oversight.

Even Technology Jobs Are Being Automated

Technology and analytics occupations stand out because the workers building digital systems are themselves exposed to automation.

AI can write code, debug software, summarize technical documents, generate queries and assist with analysis.

Yet demand for advanced technical capabilities may also increase.

McKinsey notes that US software-development job postings have recovered since early 2025, particularly for more senior and AI-oriented roles.

The likely effect is not the disappearance of technology work but a shift in the skills employers value.

Physical Jobs Are Not Immune

Manufacturing, agriculture, food service, logistics and installation also appear high on the exposure list.

AI by itself may not perform physical labor, but it increasingly works alongside robotics, computer vision, autonomous equipment and automated control systems.

That combination expands the range of tasks machines can perform.

The pace of adoption, however, may be slower in unpredictable physical environments than in purely digital work.

Automation Exposure and Head-Count Decline Are Different Metrics

One of the most important lessons from the McKinsey data is that the two percentages in the table measure different things.

The share of work hours exposed to automation describes how much of today's activity technology may be able to perform.

The decline in labor demand estimates how the number of workers required may change after economic and organizational responses.

That is why business and finance can have 62% of work hours exposed while labor demand falls by only about 18%.

The work changes more dramatically than the number of jobs.

More Than 70% of Workers May Need to Reinvent Their Roles

Occupational switching is only one form of disruption.

McKinsey estimates that more than 70% of workers could require some degree of role reinvention, meaning newly introduced tasks could account for more than 15% of their working time.

For many people, the AI transition may therefore happen without a formal career change.

Their job title remains, but their workflow does not.

Only One in Seven Workers Has an Easy Transition Path

For workers who do need to change occupations, finding another job is not necessarily enough.

McKinsey models three categories of transition.

Only about 14% have a direct pathway requiring little retraining and no significant wage loss.

About 41% face a winding path involving meaningful retraining.

Another 45% face what McKinsey calls an unpaved pathway, involving larger skill gaps, credential requirements, changes in work context or possible income loss.

The problem is therefore not simply whether vacancies exist.

Workers need realistic routes into them.

Credential Requirements Could Become a Major Barrier

Many growing occupations require licenses, certificates, training or qualifications that workers from declining industries do not currently possess.

McKinsey estimates roughly 85% of growing jobs require some form of credential or certification.

That can make a theoretically available job inaccessible in practice.

A worker may have adjacent skills but still need months of training, money for certification or relocation to another region.

The US May Need 770,000 Career Switches Every Year

Meeting the projected transition need would require about 770,000 workers per year to move between occupational groups.

Historically, the average has been about 215,000.

That means the required pace would be approximately 3.6 times the historical norm.

The United States briefly experienced a similar transition rate during the pandemic.

The difference is duration: the AI-driven transition could continue for a decade.

Skills That Complement AI Become More Valuable

A workforce operating alongside AI will need more than technical knowledge.

Problem solving, critical thinking, communication, collaboration, creativity, judgment and the ability to learn continuously become more important when routine tasks are delegated to machines.

McKinsey has argued that successful organizations will need employees who know how to work with AI, while strengthening distinctly human capabilities.

AI literacy itself is also becoming a workplace skill.

This Is a US Forecast, Not a Global Job-Loss Number

Another important limitation is geography.

McKinsey's 36 million, 11 million and 40-million-plus estimates concern the United States labor market.

They should not be presented as global forecasts.

Nor can they simply be applied to Indonesia.

Countries differ significantly in wages, sector composition, demographics, regulation, education, digital infrastructure and the economics of replacing human labor with technology.

The US study is useful for understanding mechanisms of disruption, but not for calculating Indonesian job losses.

Bottom Line

McKinsey's latest research does not forecast that AI will simply eliminate 36 million American jobs.

It estimates that AI-enabled automation and related technologies could reduce labor demand by the equivalent of roughly 36 million jobs by 2035.

About 25 million of that impact could be absorbed by growth within existing occupations.

Around 11 million workers may need to change occupations entirely in the base scenario, with a modeled range of six million to 16 million.

At the same time, economic expansion and AI-related activity could generate demand for more than 40 million jobs.

The defining challenge may therefore be less about a world without work and more about a labor market in which work moves faster than millions of people can retrain.