Will AI take your job? The answer depends on your profession

I get asked this in almost every workshop, and usually in the break rather than in front of the room. That alone tells you something: people are frightened enough to ask, and embarrassed enough to ask privately.

The answers available to them come in two kinds: "yes, retrain immediately", and "no, relax". Both were written to be shared rather than to be true.

The frame: replace, reshape, create#

Replace. Tasks with a defined output, structured input, and an error that shows itself immediately. Transcribing an invoice, converting a format, a first-pass translation, summarising minutes.

Reshape. Tasks a system does quickly at a quality that needs review. Here the job does not disappear; its contents change. Less producing, more judging. This is the largest category and the least discussed.

Create. Roles that exist because the systems exist: whoever designs the evaluation, audits the outputs, decides what must not be automated. Real examples are in jobs being born right now.

Test your own profession in twenty minutes#

This exercise beats any global report, because a report describes an average and nobody works in an average.

Step one. Write down everything you did at work last week, item by item, not as headings. "Reviewed" is not a task. "Reviewed twelve contracts for the termination clause" is a task.

Step two. Beside each: how long it took, whether the output is defined or a matter of judgement, and what happens if it is wrong.

Step three. Sort each item into the three categories.

Step four. Add up the hours per category. The proportion you get is your answer, not any article's.

What the global reports get wrong#

They measure the profession, not the person. Two accountants with identical titles can differ twofold in exposure.

They assume instant adoption. Between technically possible and actually happening sit regulation, contracts, insurance and human habit. Years, not months.

They ignore induced demand. When a task gets cheap, consumption of it sometimes rises rather than vanishing.

The answer I actually give#

When somebody asks me in the break, I do not say relax and I do not say run. I say one thing:

The risk is not that a system replaces you. It is that a colleague using one does. The difference is that the first is entirely outside your control and the second is entirely inside it.

Hence a practical recommendation: do not learn "AI" as a field. Take your most repeated and least enjoyable task and try to halve the time it costs you. Someone who has done that once understands the limits of these tools better than someone who has attended ten seminars.

This guide, profession by profession#

• Accountants
• Teachers
• Designers
• Programmers

In closing#

Nobody can tell you whether your job will disappear, but nobody knows your tasks the way you do. Spend twenty minutes on the exercise above and you will hold better information about your own future than any expert holds about you.

Try this today: write down your five most repeated tasks. If all five land in "replace", you have valuable information and time to act on it.

Common questions#

Will AI replace my job?

Most jobs will not vanish entirely, but parts of them will be automated. A job is a bundle of tasks and automation arrives task by task, so the honest answer is a proportion rather than a yes or no.

Which tasks are most exposed?

Those with a defined output, structured input, and errors that surface immediately. Tasks whose failures are expensive, or whose judgement requires human accountability, last far longer.

How do I assess my own profession?

Log a full week of tasks in detail, note each one's duration, output clarity and cost of error, then sort into replace, reshape or create. The resulting proportion beats any general report.

What is the best practical step now?

Take your most repeated, least enjoyable task and try to halve it with one tool. The real risk is not the system; it is a colleague using it while you do not.