AI and the Future of Work: How Job Roles Are Being Redefined in 2026

What research says about AI and jobs in 2026: ILO exposure estimates, WEF employer projections, Stanford's payroll data on a 19% gap for young workers, which tasks and roles are changing, and what employers and workers can do.

PUBLISHED // UPDATED
DURATION // 08 MIN READ
AI and the Future of Work: How Job Roles Are Being Redefined in 2026
FIG.01 // VISUAL DOCUMENTATION AI and the Future of Work: How Job Roles Are Being Redefined in 2026

Predictions about AI and jobs have swung between two extremes: mass unemployment or no real change. The evidence so far supports neither. Large studies find little sign of economy-wide job losses, but they do show specific pressures, especially on young workers entering occupations where AI substitutes for routine tasks. Most jobs are changing at the level of individual tasks rather than disappearing outright.

This guide summarizes what major research from the International Labour Organization, the World Economic Forum, and academic economists says about AI’s effects on work, which roles and skills are changing, and what employers and workers can do.

What the evidence says

Exposure: one in four workers

The International Labour Organization’s 2025 study, Generative AI and Jobs: A Refined Global Index of Occupational Exposure, estimates that globally, one in four workers are in an occupation with some exposure to generative AI. Key findings:

FindingFigure
Share of global employment in the highest exposure category3.3%
Highest-exposure share, women vs. men (global)4.7% vs. 2.4%
Overall exposure, low-income vs. high-income countries11% vs. 34% of employment
Most exposed occupational groupClerical occupations

The ILO stresses that most occupations consist of tasks that still require human input, so exposure usually means transformation rather than replacement. Its 2023 global analysis reached a similar conclusion: the main effect of generative AI is likely to augment rather than automate jobs, with greater impacts in higher-income countries and among women because of their representation in clerical work.

Employer expectations: churn, not collapse

The World Economic Forum’s Future of Jobs Report 2025, based on a survey of employers, projects that about 170 million jobs will be created and 92 million displaced by 2030, a net gain of about 78 million. Employers also expect around 39% of workers’ core skills to change over that period. These are employer projections, not measurements, but they point to substantial reshuffling within and between occupations.

Actual employment: young workers feel it first

The most closely watched evidence on real employment comes from Stanford’s Digital Economy Lab. In Canaries in the Coal Mine?, Erik Brynjolfsson and coauthors use payroll data to track employment since generative AI spread. Their August 2026 revision reports:

  • No evidence of widespread, economy-wide job displacement.
  • Employment of workers aged 22 to 25 in AI-exposed occupations is 19% below where it would be had it kept pace with less-exposed peers, a gap that has widened since they first documented it in August 2025.
  • Experienced workers show no comparable gap.
  • The effect works mainly through reduced hiring of young workers, not increased layoffs.
  • Declines are concentrated in occupations where AI mainly substitutes for human tasks. Where AI mainly complements workers, employment is flat or rising.

Productivity: largest gains for less experienced workers

In a study of customer-support agents, Brynjolfsson, Li, and Raymond found that an AI assistant raised productivity by nearly 14% on average, with the largest gains for newer, less experienced agents. Together with the Stanford employment data, this creates a tension: AI can help junior workers perform better, yet firms may hire fewer of them.

Tasks, not jobs

Most jobs are bundles of tasks. AI tends to affect some tasks heavily and others little:

RoleTasks AI increasingly assists or automatesTasks that remain largely human
Customer service agentAnswering routine questions, drafting replies, summarizing casesComplex complaints, judgment calls, empathy
Software developerBoilerplate code, tests, documentationArchitecture, requirements, review, accountability
Paralegal or lawyerDocument review, first drafts, research summariesStrategy, negotiation, advice, court work
Administrative assistantScheduling, form filling, correspondence draftsCoordination, discretion, relationships
Marketing specialistFirst-draft copy, variations, image generationStrategy, brand judgment, approvals

The pattern explains the Stanford finding: the routine tasks that AI handles well are often the ones junior staff used to learn on.

Roles that are growing

New and expanding roles tend to involve building, supervising, or governing AI:

  • AI operations and supervision: monitoring AI agents, handling escalations, and checking quality.
  • AI governance, risk, and compliance: impact assessments, audits, and policy work as regulations such as the EU AI Act take effect.
  • Data and evaluation specialists: preparing data, building test sets, and evaluating model performance.
  • Conversation and workflow designers: designing how AI tools interact with customers and employees.
  • Domain experts with AI skills: professionals in finance, law, medicine, and other fields who can use and critically evaluate AI outputs.

Skills gaining value

  • Judgment and verification: deciding when AI output is right, wrong, or incomplete.
  • Domain expertise: knowing enough about a field to use AI well and spot errors.
  • Problem framing: defining what needs to be done and breaking it into tasks.
  • Communication and relationships: negotiation, persuasion, care, and trust-building.
  • AI fluency: understanding what tools can and cannot do, and using them safely.

The entry-level problem

If firms stop hiring juniors because AI handles their routine tasks, they risk a future shortage of experienced staff. Organizations can address this by:

  • Redesigning junior roles around supervising, checking, and improving AI output rather than eliminating them.
  • Keeping structured mentoring so juniors still learn judgment from experienced colleagues.
  • Rotating new staff through varied tasks, including some done without AI, to build fundamentals.
  • Measuring productivity at the team level, so the value of developing junior staff is visible.

Measuring AI’s effect inside your organization

National statistics move slowly, so employers need their own evidence about how AI is changing work. A few internal measures help:

  • Hiring mix by level. Track how many entry-level versus experienced roles you open and fill each quarter in teams that use AI heavily. A falling share of junior hires is an early sign of the pattern the Stanford researchers describe.
  • Task time and quality. For key roles, sample how time is spent on major tasks before and after AI tools are introduced, and track error and rework rates alongside speed.
  • Internal mobility. Count how many employees move into new or redesigned roles, such as AI operations or quality review, rather than leaving.
  • Training completion and use. Record not just course completion but whether trained staff actually use AI tools in their work.
  • Employee feedback. Regular short surveys on workload, job quality, and where AI helps or hinders give context the numbers miss.

Reviewing these measures with HR, finance, and team leaders at least twice a year turns broad predictions into decisions about hiring, training, and role design.

What employers can do

  1. Map tasks, not jobs. Identify which tasks in each role AI can assist or automate.
  2. Decide where AI substitutes and where it complements. The Stanford data suggests complementary uses are better for employment.
  3. Invest in reskilling existing staff who already understand the business.
  4. Involve workers in redesigning roles; they know where AI helps and where it causes problems.
  5. Monitor quality, not just speed, after introducing AI.
  6. Communicate honestly about plans, which reduces resistance and turnover.

What workers can do

  • Build deep knowledge in a domain and learn to apply AI within it.
  • Practice checking AI output critically rather than accepting it.
  • Seek roles and tasks where AI complements your work.
  • Develop communication and coordination skills that are hard to automate.
  • Keep learning as tools change.

For related analysis, see our articles on AI in customer support, measuring AI ROI, and Gartner’s finding on role redesign for AI agents.

Bottom line

The best available evidence shows AI transforming work task by task rather than eliminating jobs across the economy. The ILO estimates one in four workers globally are in occupations with some generative AI exposure, mostly in ways that augment work. The WEF’s employer survey expects large churn but a net gain in jobs by 2030. Stanford’s payroll research finds no widespread displacement, but a growing 19% employment gap for young workers in AI-exposed occupations, driven by reduced hiring where AI substitutes for tasks. Employers that redesign junior roles, invest in reskilling, and favor uses that complement workers, and workers who pair domain expertise with AI fluency and judgment, are best placed for the transition.

Sources