McKinsey: AI and robots could technically automate 57% of US work hours

A new report from the McKinsey Global Institute published in November 2025 estimates that currently demonstrated AI and robotics technologies could, in theory, automate activities accounting for about 57 percent of US work hours today. At the most automatable end of the occupational spectrum, roles with the highest potential for automation by agents or robots make up about 40 percent of total jobs.

The report, titled Agents, robots, and us: Skill partnerships in the age of AI, is careful to call this a measure of technical potential, not a jobs forecast. But the numbers are still the kind that reshape public debate. And they arrive at a moment when the distance between technical potential and economic reality is shrinking faster than most policy institutions are ready for.

McKinsey breaks the automation story into two distinct technological forces. AI agents — software systems that automate non-physical work — could perform tasks occupying 44 percent of US work hours today. Physical robots could handle another 13 percent. Together, they cover more than half of everything currently done in the American economy.

The 40 percent jobs figure is the one that will dominate headlines. These are roles concentrated in legal and administrative services, plus physically demanding jobs like driving and machine operation. McKinsey is careful to note that many of these jobs will evolve rather than disappear. Tasks get redistributed between humans, agents, and robots, with people still needed to guide, supervise, and verify.

That framing is optimistic. It is also incomplete.

The report’s own data suggests that the pressure on those 40 percent of jobs is not evenly distributed. Exhibit 2 in the report maps the workforce against a 2x2 matrix of people, agents, and robots. Non-physical work accounts for about two-thirds of US work hours. The most automatable activities represent about 40 percent of total US wages and span roles from education and healthcare to business and legal.

What McKinsey calls “people-centric” roles — many healthcare and building maintenance jobs — remain largely human-led. But “agent-centric” and “robot-centric” roles overlap strongly with that 40 percent figure. The report classifies roughly 800 occupations into seven archetypes, from people-centric to agent-centric to robot-centric and mixed. The agent-centric and robot-centric categories are where the transformation happens fastest.

The skill story is more nuanced than a simple obsolescence narrative. McKinsey finds that more than 70 percent of the skills sought by employers today are used in both automatable and non-automatable work. Roughly 72 percent of skills are required both for work that could be done by AI and for work that must be done by people.

To track this, the report introduces a new “Skill Change Index” — a time-weighted measure of automation’s potential impact on each skill used in today’s workforce. Digital and information-processing skills are expected to see the biggest shifts. Assisting and caring skills see the least change. A core set of eight “high-prevalence skills” — including communication, management, problem-solving, leadership, customer relations, and writing — remain essential across industries. They sit in the middle of the distribution, likely to be reshaped rather than eliminated.

The most striking data point in the report is not about automation potential at all. It is about how fast the labor market is already moving.

Drawing on US job-posting data, McKinsey finds that demand for AI fluency — the ability to use and manage AI tools — has grown sevenfold in two years, faster than for any other skill in US job postings. The number of workers in occupations where AI fluency is explicitly required has risen from around 1 million in 2023 to about 7 million in 2025.

That is a sevenfold increase in two years. For comparison, demand for technical AI skills is rising more slowly. Three-quarters of AI skill demand is currently concentrated in three occupational groups: computer and mathematical, management, and business and financial operations. Construction, transportation, and food service show relatively little demand so far.

At the same time, mentions of skills such as general science and research, and writing and editing, are declining in job listings — even though, as the report notes, these skills remain essential for much of the workforce. The market is signaling a shift that may outpace the actual capability of the technology.

McKinsey’s economic argument is where the report gets most interesting. In its midpoint adoption scenario, AI-powered agents and robots could generate about $2.9 trillion in US economic value per year by 2030. But that number comes with a condition: only if organizations prepare their people and redesign workflows, rather than individual tasks, around people, agents, and robots working together.

About 60 percent of that value sits in sector-specific workflows — supply chain management in manufacturing, clinical diagnosis in healthcare, risk management in finance. The rest sits in cross-cutting functions like IT, finance, and administration.

The report includes case studies that make the abstraction concrete. A global tech company uses multiple AI agents to qualify sales leads and schedule meetings, freeing human specialists to negotiate and build relationships. A large utility uses conversational agents to handle common customer queries, cutting the average cost per call by about 50 percent. A pharmaceutical company deploys generative AI to draft clinical study reports, reducing touch time for first human-reviewed drafts by nearly 60 percent and cutting errors by around 50 percent. A regional bank uses AI agents to accelerate code migration, with engineers shifting from manual rewriting to planning, orchestration, and testing.

McKinsey’s $2.9 trillion prize comes with a condition: only if organizations redesign workflows, not just tasks, around people, agents, and robots working together.

Across these examples, managers move away from routine supervision and towards orchestrating systems in which people, AI agents, and robots collaborate. New performance metrics and AI-related skills become the norm.

The report is not a forecast of mass unemployment. The authors write explicitly that the 57 percent and 40 percent figures reflect technical potential, not a prediction. They argue that AI will not make most human skills obsolete, but it will change how they are used.

That is the right caveat. It is also the wrong comfort. The gap between technical potential and economic adoption has historically been wide. But the speed of AI deployment in enterprise workflows — particularly agent-based systems — is compressing that gap faster than any previous automation technology. The sevenfold increase in AI fluency demand in two years is not a lagging indicator. It is a leading one.

The question the report raises but does not answer is whether the institutions that manage labor transitions — retraining programs, safety nets, education systems, collective bargaining — can operate at the same speed as the technology. McKinsey’s $2.9 trillion prize depends on workflow redesign. But workflow redesign depends on people who can do the redesigning. And those people are the same ones whose skills are shifting fastest.

The report gives policymakers, employers, and workers a clearer view of where the pressure lands first. The 40 percent figure is not a prediction of job losses. It is a map of where the ground is about to move.