Will AI take your job? The honest answer from research is more nuanced than headlines suggest. Studies measure how exposed occupations are to AI — how many of their tasks AI could perform or speed up. High exposure can mean automation, but it can also mean augmentation: the same job, done differently and perhaps more productively.

How exposure is measured

Researchers break each occupation into tasks using databases such as the US O*NET, then estimate which tasks language models could do or accelerate significantly. An occupation’s exposure reflects the share and importance of exposed tasks.

Key findings

OpenAI and University of Pennsylvania researchers (2023), in the paper GPTs are GPTs, estimated that around 80% of the US workforce could have at least 10% of their tasks affected by large language models, and around 19% of workers could see at least half of their tasks affected. Exposure was higher for higher-wage, more educated occupations — the reverse of earlier automation waves that mainly affected routine manual work.

The International Labour Organization (2023) analyzed jobs globally and concluded that the most likely effect of generative AI is augmentation rather than automation for most occupations. Clerical support work showed the highest exposure to automation, and because women are overrepresented in clerical jobs in many countries, they were found to be disproportionately exposed. Exposure was also higher in high-income countries.

Highly exposed task areas

  • drafting and editing text;
  • summarizing documents;
  • answering routine customer questions;
  • data entry and processing;
  • translation;
  • basic coding and analysis;
  • research and information gathering.

Less exposed work

  • physical and manual work in unpredictable environments (trades, care, construction);
  • work requiring physical presence and dexterity;
  • roles centered on human relationships, trust and persuasion;
  • high-stakes judgment with accountability.

Exposure is not destiny

Whether exposure leads to job losses depends on:

  • costs and reliability of AI for the task;
  • regulation and liability in the sector;
  • demand: if AI makes a service cheaper, demand may grow, offsetting losses;
  • organizational choices: whether firms use AI to cut staff or to expand output and quality;
  • worker skills and training.

Field studies show gains for workers using AI — for example, support agents with an AI assistant resolved more issues per hour, with the largest gains for less experienced staff, according to a widely cited NBER study.

What workers can do

  • Learn to use AI tools in your own field.
  • Strengthen skills AI complements: judgment, communication, domain expertise, relationship building.
  • Focus on tasks where accountability and trust matter.
  • Watch how your industry adopts AI and look for roles that combine domain knowledge with AI fluency.

How to assess your own role

List your weekly tasks and estimate the share that involves drafting, summarizing, routine analysis or information lookup — the tasks most exposed. Then identify tasks requiring judgment, relationships, physical presence or accountability. Building strength in the second group is the best hedge.

What to watch next

  • Real employment data. Early-career hiring trends in highly exposed occupations are a key indicator to follow.
  • Agents. AI that completes multi-step tasks could raise exposure for roles involving coordination and back-office work.
  • Productivity statistics. Whether gains seen in field studies appear in national productivity data.
  • Reskilling programs. Government and employer initiatives to retrain workers in exposed roles.
  • Workplace rules. Collective agreements and regulations on how AI may be used to monitor or manage workers.

Key takeaways

  • Exposure measures how many of a job’s tasks AI could perform or speed up, not certain job loss.
  • Studies found generative AI exposure is higher for many higher-wage, educated occupations.
  • The ILO found augmentation more likely than automation for most jobs, with clerical work most at risk.
  • Outcomes depend on cost, regulation, demand and organizational choices.
  • Building AI fluency plus judgment and domain expertise is the best personal strategy.

Frequently asked questions

Will AI cause mass unemployment?

Most economists expect significant change in tasks and jobs rather than mass unemployment, but the transition could be disruptive for specific occupations and regions.

Which jobs are safest?

Jobs combining physical work in varied environments, human relationships and accountable judgment are generally least exposed to current generative AI.

Should I learn to code given AI coding tools?

Programming fundamentals remain valuable for understanding and supervising AI-generated code, though the mix of skills needed is changing.