Will AI replace my job

I would not promise that AI will leave your job untouched. I would also question anyone who claims to know your future from your job title alone. The useful starting point is the work you actually do, the parts a system can perform and the decisions your employer makes after that.

A technical demonstration shows that something can be done under particular conditions. A staffing decision adds questions about cost, demand, reliability and who will handle the work the system leaves behind. Your concern about employment is legitimate even when the demonstration is incomplete.

Exposure is an estimate of tasks AI could affect

The ILO–NASK global index published in May 2025 estimates that one in four workers worldwide is in an occupation with some exposure to generative AI. The researchers combine task information, human assessments and AI predictions to assess potential exposure. They are not counting people who have lost their jobs.

A task may be suitable for AI assistance while other duties continue to require a person. The index identifies clerical occupations as especially exposed and judges transformation to be the most likely overall effect. That is a global assessment, not a guarantee that every affected worker keeps their position.

ILO economist Janine Berg put the communication problem plainly: What we need is clarity and context. Her statement accompanied the index's release.

Open up an ordinary working week

Take a hypothetical customer-service job. The work may include finding an account, drafting a reply, correcting a record, calming an upset customer and deciding when a complaint needs escalation. Those activities have different information requirements and consequences. Treating the job as a single block hides the decisions that determine whether automation is useful.

Write down what takes your time over a working week. For each task, ask whether AI can complete it to the required standard, whether someone must check it and whether it involves an action the system is authorised to take. Mark what you have tested separately from what a product demonstration suggests.

Some duties may be unsuitable for the tools available to your organisation. Others may be technically possible but uneconomic once checking and integration are included. Keep those reasons separate: a current barrier can change, and a policy choice deserves an explicit discussion.

Less work per case does not settle headcount

Suppose a team can handle each routine case faster. Management could serve more customers, reduce waiting time, move staff to difficult cases or reduce staffing. These are possible responses, not forecasts. The result depends partly on how much work exists and what the organisation chooses to prioritise.

This is why I dislike reassurance that stops at “AI will change tasks”. A smaller collection of tasks can still mean fewer jobs. The honest discussion includes that possibility, the evidence needed before changing staffing and what support affected employees will receive.

Prepare for a specific change

Choose one task that is already changing and learn how to perform, check and explain it with AI. Keep evidence of the result and the time required, including corrections. Then identify the adjacent work you could take on if the routine part becomes smaller. This is preparation for a concrete change rather than a promise of permanent protection.

Your employer has responsibilities in that conversation too. Ask what the organisation expects to automate, what it has tested, how staff will participate and when it will make decisions. Training without a clear account of the intended change leaves people guessing.

You deserve to know what is being proposed while there is still time to influence it.