The most urgent question about artificial intelligence is not philosophical. It is practical.

Who will still have a job?

Economists, technologists, and policymakers are trying to answer that question at the same time that the technology itself is still evolving. The result is a strange mixture of confidence and uncertainty. Studies suggest that many occupations may change dramatically under the influence of AI. Yet the same research also warns that our predictions about technological change are famously unreliable.

Still, a few patterns are beginning to emerge.

The New Shape of Automation

For most of the past century, automation threatened physical labor first. Factory workers, agricultural workers, and tradespeople were the ones most likely to see machines take over their tasks.

Artificial intelligence reverses that pattern.

The occupations most exposed to AI are often white-collar roles that revolve around language, analysis, or digital communication. Researchers studying more than 350 occupations have found that many tasks performed by writers, programmers, translators, marketers, and customer service workers overlap heavily with capabilities already demonstrated by modern AI systems.

In other words, the same systems that write emails, summarize documents, and generate computer code are touching the core tasks of many professional jobs.

By contrast, work that requires physical presence or complex real-world interaction remains harder to automate. Firefighters, nurses, and fast-food workers appear less exposed to direct AI substitution, not because the work is simple, but because it requires coordination in the physical world.

This is one of the paradoxes of the AI era: some of the most prestigious knowledge jobs may be easier to automate than certain hands-on roles.

Exposure Is Only Half the Story

But exposure to AI does not necessarily mean disappearance.

Researchers at GovAI and the Brookings Institution added a second dimension to the analysis: adaptability. They examined whether workers in different occupations would likely be able to shift into other jobs if AI disrupted their current one.

They considered factors such as:

• education and transferable skills
• age
• personal savings
• access to large job markets
• diversity of prior work experience

The results revealed an unexpected split between two groups that, on paper, face similar levels of AI exposure.

Take web designers and secretaries.

Both professions perform many digital tasks that AI systems increasingly handle well: drafting text, organizing information, formatting documents, and interacting with digital tools. Yet the researchers concluded that web designers are far more likely to adapt successfully if their jobs change or disappear.

Why?

Because their technical skills transfer more easily into adjacent fields such as software development, product design, or digital marketing.

Secretaries, by contrast, often work in narrower administrative roles with fewer clear pathways into new professions. The same technological tools that promise efficiency improvements could also compress the number of administrative jobs needed inside organizations.

The Gender Divide

One of the most striking findings is that the workers most vulnerable to AI disruption are disproportionately women.

Clerical and administrative jobs — including secretaries, administrative assistants, and similar roles — have historically been female-dominated occupations. Researchers estimate that women make up roughly 86 percent of the workers in the group that is both highly exposed to AI and least adaptable to job changes.

This pattern echoes earlier technological shifts.

Throughout the twentieth century, office technologies were often introduced with promises that they would elevate administrative workers into higher-level roles. In practice, many organizations simply used those tools to increase productivity expectations without significantly improving wages or career mobility.

The introduction of spreadsheets, word processors, and office software rarely eliminated administrative work entirely. Instead, it allowed fewer workers to perform the same tasks for larger organizations.

Artificial intelligence may accelerate that dynamic.

The Prediction Problem

Despite the growing volume of research, economists repeatedly caution that forecasts about AI and employment should be treated with humility.

History is full of confident predictions that turned out to be wrong.

Early automation studies suggested bank tellers would vanish after the spread of ATMs. Instead, the number of bank branches increased and the role of tellers evolved toward customer service and financial guidance.

Medical experts once predicted that computer diagnostics would sharply reduce the need for radiologists. Instead, imaging demand expanded faster than automation.

Even earlier technological revolutions show the same pattern. The electrification of buildings created the now-extinct job of elevator operator, which disappeared decades later when automatic buttons replaced human control.

Technological progress rarely eliminates work in neat, predictable ways. It tends to rearrange it.

The economist John Maynard Keynes once predicted that technological efficiency would eventually produce a fifteen-hour workweek. Nearly a century later, productivity has soared — but most people still work full schedules.

Contradictory Evidence

Current data about AI’s effect on employment is already producing conflicting interpretations.

Some studies suggest that young workers in highly exposed fields such as software development and customer service are seeing fewer opportunities as AI tools become widely adopted.

Other analyses find the opposite: young workers in those fields may be benefiting because familiarity with AI tools makes them more productive and therefore more attractive to employers.

Meanwhile, some central banks believe that large-scale job displacement from AI may take decades, while prominent technology executives warn that disruption could arrive much sooner.

In truth, both outcomes may occur simultaneously in different industries.

A More Complicated Future

Two points of agreement are emerging across most research.

First, there is no clear evidence yet that AI is eliminating jobs across the entire economy.

Second, the first wave of disruption appears to be targeting white-collar work, not the manual labor traditionally associated with automation.

But the most important lesson from history may be simpler.

Technology rarely eliminates the need for human work altogether. Instead, it redistributes opportunity unevenly. Some professions shrink. Others evolve. Entirely new ones appear that no one previously imagined.

The invention of the smartphone created occupations — social media managers, app developers, digital influencers — that barely existed twenty years ago.

Artificial intelligence will likely produce similar surprises.

The challenge for workers is not predicting the future with certainty. It is cultivating the skills that make adaptation possible when the future arrives.

In that sense, the real divide in the age of AI may not be between jobs that survive and jobs that disappear.

It may be between workers who can pivot — and those who cannot.