Will AI take your job? Most probably not in the way you think.

If you spend enough time online, AI is either about to make all of us unemployed or make all of us ten times more productive by Tuesday. The research is considerably less dramatic - and considerably more interesting.

The latest evidence suggests that AI is unlikely to sweep across the labour market evenly. Some jobs are genuinely vulnerable, many more are likely to be reorganised, some could grow, and a surprisingly large part of the workforce may see much less immediate change.

Which parts of my job are becoming easier for AI to do, and which parts are becoming more valuable because AI cannot?

First, about that 18% number

In April 2026, OpenAI published a labour-market framework covering 921 occupations and approximately 148 million US jobs. Its estimate: 18% of jobs face relatively high automation risk.

That is not a prediction that 18% of people will lose their jobs. It identifies occupations where enough work can potentially be automated that employment pressure becomes more plausible. That distinction matters.

So which jobs are actually in the 18%?

Think less about job titles for a moment and more about the work underneath them. AI is particularly good at work that involves:

• processing standardised information
• producing predictable text
• checking or classifying information
• following documented procedures
• converting one form of digital information into another
• answering repeated, structured questions

That puts jobs such as data-entry clerks, telemarketers, proofreaders, bookkeeping roles and some administrative jobs closer to the front of the automation queue. Exposure is also spreading into professional roles that happen largely on screens.

AI can do part of my job is not the same as my job disappears

A lawyer does not exist simply to draft text. A teacher does not exist simply to produce lesson plans. A financial analyst does not exist simply to analyse a spreadsheet, and a software developer does not exist simply to type code.

AI can take over meaningful pieces of the work, but someone still needs to exercise judgement, deal with unusual cases, take responsibility, understand context, manage relationships and notice when the answer is wrong.

The job survives. The task mix changes.

The latest real-world data is surprisingly reassuring

By May 2026, 45% of US workers said they used generative AI for their jobs, and in more than 80% of occupations at least one in five workers was using AI. So yes, AI is already everywhere. But the depth of that use tells a different story.

Fewer than 3% of individual job tasks were being regularly performed with AI by at least half of the workers who perform those tasks. AI is appearing in a lot of jobs without taking over most of any one job.

Some work is barely changing at all

AI works extremely well in the digital world. It remains much less useful when the work itself has to happen in the physical world. It can help an electrician write an estimate, but it cannot currently crawl underneath your sink and fix the pipe.

Are jobs going to disappear?

Yes, some already are declining. But new jobs are being created at the same time. The World Economic Forum estimates that major economic, demographic, technological and environmental changes could create 170 million jobs globally while displacing 92 million between 2025 and 2030.

The labour market is not moving from humans to AI. It is moving from one way of dividing work between humans and technology to another.