Will AI Replace My Job? A Task-by-Task Way to Find Out

Start with tasks, not headlines

A job title is a container. Inside it are many different responsibilities: researching, drafting, checking, persuading, deciding, coordinating, handling exceptions, building trust and sometimes doing physical work. AI does not affect all of those responsibilities equally.

That is why a headline such as “AI will replace accountants” or “AI will replace marketers” is too blunt to be useful. Two people with the same title can have very different exposure depending on what they actually do every week.

The International Labour Organization’s 2025 update uses task-level analysis and finds that one in four workers globally are in occupations with some degree of generative-AI exposure. Its central conclusion is more nuanced than mass replacement: because most occupations still contain tasks requiring human input, transformation is more likely than wholesale redundancy.

A five-question task test

Take one real responsibility from your job and examine it on its own. You are not trying to predict the future perfectly. You are trying to identify where change is most likely to reach your work first.

• Is the task repetitive enough that the steps are mostly the same each time?
• Is it rules-based, with clear inputs, outputs or decision criteria?
• Is most of the work digital and information-heavy rather than physical?
• Can the output be checked quickly by a competent human?
• Does success depend heavily on judgment, accountability, relationships, context or trust?

Sort your work into automate, augment and deepen

Tasks that are repetitive, digital and easy to verify are usually the first candidates to automate or partially automate. That does not mean you should remove yourself from the process. It means you should test whether AI can reduce the low-value steps safely.

Tasks that benefit from speed but still need your review belong in the augment category. Research, first drafts, summarisation, analysis preparation and option generation often sit here. Your value shifts from producing every first draft manually to directing, checking and improving the work.

Then identify what to deepen: the work where your judgment, relationships, responsibility, domain knowledge or ability to handle uncertainty matters most. These are not magically immune to AI, but they are harder to compress into a clean prompt-and-output loop.

Exposure is not the same as unemployment

This distinction matters. A task can be exposed to AI without the person doing the job becoming unnecessary. The OECD has found that many workers in AI-exposed occupations are unlikely to need specialist machine-learning skills, even though the tasks they perform and the skills demanded of them may change.

The World Economic Forum’s Future of Jobs Report 2025 estimates large labour-market churn by 2030: 170 million roles created and 92 million displaced across major macrotrends, including technology. Those figures describe a changing labour market, not a simple one-way story in which every technology-exposed worker loses a job.

What to do this week

Choose five responsibilities you perform regularly. Score each one for repetition, rules, data dependence, physical presence and human judgment. Pick one high-exposure task and test a controlled AI-assisted workflow around it. Keep a human review step. Measure time saved, error rate or quality before and after.

Then document what changed. The goal is not to become the person who talks most about AI. It is to become the person who can redesign real work responsibly and show the result.