When people ask which AI jobs are at risk, they often imagine a list of occupations that will disappear. That may be the wrong unit of analysis.
Jobs are bundles of tasks. AI can automate some parts of a role, accelerate others and leave the most human parts largely untouched. That means the first jobs to change may be the ones with the highest concentration of language, analysis and repeatable digital work, not necessarily the jobs that look most “routine” from the outside.
AI affects tasks before it eliminates occupations
McKinsey’s work on generative AI has emphasized that the technology increases the automation potential of activities involving expertise, collaboration and decision-making, categories that were previously harder to automate.[1]
That is a significant departure from older automation. Industrial automation was strongest where work was repetitive and physical. Generative AI reaches directly into knowledge work because so much of that work is expressed through text, documents, code and digital systems.
White-collar work is unusually exposed
Legal research, financial analysis, marketing, software development, customer service and administrative work all contain large amounts of information processing. AI can draft, summarize, classify, compare and retrieve that information much faster than a person working manually.
This does not mean those professions vanish. It means the balance of work inside them can change quickly.
McKinsey has estimated that generative AI could substantially increase the share of work hours with automation potential in occupations involving expertise, including roles that were relatively insulated from earlier waves of automation.[2]
The junior layer may change first
Many professional careers have historically trained people through repetitive entry-level work: reviewing documents, preparing first drafts, compiling research, cleaning data or producing routine analysis.
Those are exactly the kinds of tasks AI can often assist with. That creates an uncomfortable organizational question. If companies automate the work that used to train junior employees, they need a new way to develop future senior employees.
AI therefore changes more than headcount. It can change the apprenticeship structure of a profession.
Software work is a useful preview
Coding has become one of the clearest examples of AI-assisted work. Anthropic’s research on Claude Code found that people across occupations can use coding agents successfully, while domain expertise still improves outcomes.[3]
The implication is not that expertise stops mattering. It may matter differently. The person who understands what should be built can delegate more of the implementation while remaining responsible for direction and judgment.
Physical jobs may change more slowly, but they are not untouched
AI’s direct impact is currently strongest where work is already digital. Construction, logistics, healthcare and manufacturing still involve physical constraints that language models alone cannot solve.
But robotics, computer vision and intelligent scheduling are expanding the reach of automation beyond the screen. The World Economic Forum’s 2025 research on major global job families argues that AI, robotics, energy and network technologies will affect a much wider share of the workforce than knowledge work alone.[4]
New jobs emerge around implementation
Every technology wave creates work around adoption. AI is doing the same. Companies need people who can redesign workflows, evaluate tools, connect systems, manage data, establish controls and train employees.
That is why “AI jobs” should not be interpreted only as jobs at AI companies. Many of the opportunities will exist inside ordinary businesses trying to apply AI to a specific industry.
Skills are becoming more important than titles
A job title can stay the same while the work underneath it changes dramatically. A marketer may still be a marketer, but spend less time drafting variants and more time deciding positioning. An analyst may produce fewer manual summaries and spend more time interrogating assumptions.
The useful question for workers is therefore: which parts of my job are becoming easier to automate, and which parts become more valuable when that happens?
The first advantage may go to people who can delegate well
AI agents are increasingly able to take on longer sequences of work. OpenAI has reported growing use of coding agents for tasks estimated to take people an hour or more, including among non-developers.[5]
That suggests a new professional skill: defining work clearly enough for an AI system to execute it, then evaluating the result. The person who can frame the problem, provide context and recognize a bad answer may become more productive than someone who simply performs every step manually.
AI is changing jobs before it replaces them
The labor-market impact will not be uniform, and forecasts should be treated as scenarios rather than certainties. Adoption depends on cost, regulation, organizational design and whether the technology actually works well enough in real settings.
But the direction is becoming clearer. AI is moving into the everyday tasks that make up professional work. The first transformation is likely to be inside jobs, one responsibility at a time.
Sources
[1] McKinsey & Company, “Gen AI’s automation boost”
[2] McKinsey Global Institute, “Generative AI and the future of work in America”
[3] Anthropic, “Agentic coding and persistent returns to expertise”
[4] World Economic Forum, “Jobs of Tomorrow”
[5] OpenAI, “How agents are transforming work”
Build your AI fluency.
In the AI Era™ Professional Learning is a five-level certification pathway designed to help professionals move from understanding AI to applying it with greater confidence, productivity, and sophistication.
