AI is more likely to change many jobs than erase every exposed occupation at once. That does not make the risk harmless. Tasks can disappear, entry-level routes can narrow, workloads can intensify, and worker control can weaken even when the job title survives.
Exposure is not the same as replacement.
A job is a bundle of tasks, relationships, responsibilities, and decisions. A model may perform some tasks while failing at the context, trust, physical work, or accountability around them. That is why estimates of "jobs exposed to AI" should not be read as a list of jobs scheduled for deletion.
The International Labour Organization's 2025 global index found that one in four workers is in an occupation with some generative AI exposure. Its central conclusion was that job transformation is more likely than complete redundancy because most occupations still include tasks that require human input.
workers worldwide are in occupations with some generative AI exposure, according to the ILO. Exposure measures potential task overlap, not a guaranteed layoff.
Why workers are still right to worry.
Keeping a job title does not guarantee keeping the same pay, autonomy, workload, or path into the field. If AI absorbs routine work, the remaining human work may become more complex without better compensation. If entry-level tasks disappear, new workers may lose the place where they learn judgment and build experience.
A 2026 ILO review of early evidence found that large-scale displacement remained limited, while warning about inequality, early-career opportunity, worker autonomy, and job quality. That is a more useful picture than either extreme claim: "nothing will change" or "every job is gone."
Which jobs have more exposure?
Clerical occupations remain among the most exposed because many of their tasks involve text, records, scheduling, classification, and other digitized information. The ILO also found increasing exposure in highly digitized cognitive work, including some media, software, finance, and professional tasks.
Exposure can look different inside the same occupation. A journalist may use a tool to transcribe an interview while rejecting it as a substitute for reporting. A designer may automate file preparation while protecting the creative decision. A customer service worker may receive useful search support or be forced to monitor several automated conversations at once.
The question is not only whether the job survives. It is whether the worker still has a say.
What should workers ask when AI arrives?
- What problem is the tool supposed to solve? Ask for a specific use, not a vague promise of efficiency.
- Which tasks and decisions will change? Separate drafting, recommendation, monitoring, scoring, and final authority.
- What worker data will be collected? Performance monitoring and model training should not be hidden inside a productivity feature.
- Who can challenge the system? There should be a named person, a record of important decisions, and a real appeal route.
- Who receives the gains? If output rises, ask what changes in staffing, pay, workload, training, and ownership of the work.
Human oversight must include authority.
A person cannot meaningfully oversee a system if they are punished for slowing it down, cannot see why it made a recommendation, or do not have permission to override it. "Human in the loop" is not a safeguard by itself.
Employers should document the boundary: what the tool may do, what it may not decide, when a human review is required, and who is responsible when it fails. Workers affected by the system should have a role in setting and revising those limits.
What can one worker do?
Start by asking for the policy in writing. Keep examples of changed duties, error correction, new monitoring, or unpaid extra work. Compare notes with colleagues. If a decision affects pay, evaluation, discipline, hiring, or termination, ask how a person can review and appeal it.
No individual worker should have to solve a structural problem alone. The point of asking clear questions is to turn a vague technology announcement into a visible workplace decision.
Sources and further reading.
- 01International Labour Organization: Generative AI and Jobs, 2025 update
- 02International Labour Organization: Impact of GenAI on jobs, productivity and work organization, 2026 review
- 03Pew Research Center: Workers' views of AI use in the workplace