McKinsey on Agents and Robots

The Investment Hurricane Is Real

McKinsey published a 60-page report dissecting how jobs, skills, and workflows are changing. The report moves past generic alarmism and attempts to answer what professionals should actually do next. For anyone in professional services, the conversation about Generative AI (GenAI) is no longer optional.

The investment numbers are staggering. Estimates place AI investment moving from $250 billion toward $2.5 trillion and climbing. McKinsey noted that demand for AI fluency, the ability to use and manage AI tools, has grown sevenfold in two years, faster than any other skill in United States job postings 1. Anyone in professional services, whether attorneys, consultants, bankers, or professors, has not gone a day without the topic of AI arising with customers, suppliers, and students.

We Are Not Luddites

The technology explosion in computing, storage, and broadband has delivered enormous productivity gains. Schumpeter's creative destruction remains a force. The benefits are undeniable and personal. Home broadband at 600 megabits costs roughly $70 a month. A decade ago, that speed was unavailable at any price. Smartphone cameras now far surpass the digital cameras from a decade ago.

GenAI is part of the professional workflow now. It is fait accompli. The question is no longer whether AI belongs in professional services but how professionals will adapt their work around it.

AI as Alien Intelligence

Yuval Noah Harari called AI an alien intelligence. It can reason and perform tasks previously thought to require human cognition. The unsettling reality is that GenAI is the worst it will ever be today. Unlike other technologies, AI improves through use. Feedback loops, data generation, and iteration make it stronger continuously. It can even generate synthetic data to train itself.

GenAI functions as a relentless grinder. It handles keyboard-based work, research, administration, emails, confirmations, invoices, and spreadsheet crosswalks. The volume of white-collar work that falls into this category is enormous. PwC research found that AI-exposed industries are experiencing nearly five times higher labor productivity growth than non-exposed sectors 2.

The 57 Percent Automation Figure

McKinsey noted that GenAI, combining agents and robotics, could theoretically automate about 57 percent of current United States work hours. Non-physical work represents roughly two-thirds of American work hours. About two-thirds of that non-physical work could be heavily automated. Think of all the research, administration, emails, calls, confirmations, invoices, and spreadsheet crosswalks that white-collar professionals perform daily.

Physical activities prove harder to automate, which is ironic. Robots have improved, but they lack the motor skills, dexterity, situational awareness, and battery life needed for real-world deployment. Physical automation accounts for roughly 13 percent of the theoretical total. The combination of 44 percent agentic automation and 13 percent robotic automation reaches the 57 percent figure.

Adoption Takes Time

Adoption does not happen overnight. Different industries and companies adopt at different rates. AI-native entrants will move aggressively. The vast majority of companies will take their time, wait and see, and make incremental changes only.

McKinsey observed that electricity took more than thirty years to spread. Industrial robotics followed a similar multi-decade path. As recently as 2023, only about one in five companies ran most of their applications in the cloud, despite the technology being widely available since the mid-2000s. There are many laggards. The Penn Wharton Budget Model projects that AI's impact on total factor productivity growth remains small today but will accelerate through the early 2030s 3.

People, Agents, and Robots

The claim that AI is coming for jobs is a gross generalization. Which jobs? When? Why? McKinsey analyzed 800 occupations and separated them into physical and non-physical components. This analysis produced eight segments of work with varying degrees of exposure.

People-centric roles like nurses, psychologists, and firefighters face the least change. People-agent roles like sales representatives and human resources specialists see moderate disruption. Agent-centric roles like accountants, lawyers, and software developers face more significant change. People-robot roles like insulation installers and drywallers see moderate physical automation. Robot-centric roles like pick and pack logistics and welding face more immediate turbulence.

The pattern is clear. The more judgment, trust, and human context a role requires, the slower the disruption. People-centric roles are the safest. Agent-centric and robot-centric roles face the most immediate turbulence.

600 New GenAI-Adjacent Skills

McKinsey examined 6,800 skills across 11 million job postings. Since 2023, more than 600 new skills have emerged in demand. That represents a 10 percent increase in new types of skills, and they are largely GenAI-related.

McKinsey calls these GenAI-adjacent capabilities. Roles are mutating quickly. The idea of one-time reskilling is simplistic. Professionals need highly transferable skills they can apply, learn, morph, readapt, and reuse. Adaptability outperforms specialization when the ground is moving. Think of skills as Lego bricks that get restacked and rearranged constantly.

Employers increasingly seek AI-adjacent capabilities such as process optimization, quality assurance, and teaching. These skills serve to redesign work with AI, supervise and verify AI systems, or train people to use them.

Value Migrates Upward

All professionals will gravitate toward more value-added work. Instead of analyzing data, professionals will interpret findings. Instead of forecasting demand, they will create scenario plans. The machine handles the preparation work, giving professionals more time to hone their craft, perform quality control, and engage with stakeholders.

This shift resembles a maturity model. Professionals will perform level four and level five work rather than level two work. The tedious, mechanical portions of professional service give way to interpretation, judgment, and strategic synthesis.

A Thought Experiment for Professionals

Consider what work is less likely to go to the machine. Curating content and determining which materials are relevant, useful, and authoritative requires human judgment. Facilitating live discussion and creating a safe space for constructive tension and synthesis demands presence. Qualifying a potential client project for scope, budget, and fit requires contextual understanding.

Following up with a student who misunderstood a concept during class discussion requires empathy. Holding an office hour with someone weighing the merits of a specific internship requires lived experience. These tasks share a common thread. They require judgment, trust, and human context, the three qualities that slow disruption.

What to Do Next

Skim through the McKinsey report. Download it, upload it into an AI assistant, and ask it to help segregate your specific work. Spend less time on keyboard work. Spend more time on judgment work. Invest in transferable skills. Help with decisions rather than research. The professionals who thrive will be those who use AI to eliminate drudgery and elevate themselves toward the work that machines cannot do.

• 1McKinsey
• 2PwC
• 3Penn Wharton Budget Model