Where Did That Number Come From?
You've probably seen a headline like this: "40% of jobs hit by AI!" Maybe it made your stomach drop a little.
So here's a question for you. Who said it? When? And what did they actually count?
If you can't answer those three things, you can't really judge the number. In Level 1 we saw that AI changes tasks first and whole jobs only sometimes. Today we use that idea to read the big studies properly. Think of it as reading the label on a food packet before you decide what's in it.
What Does "Exposure" Actually Mean?
This word does a lot of work in these reports, so let's get it straight.
When researchers say a job is "exposed" to AI, they usually mean that a good share of its tasks overlap with what AI can do. That's all. It's a measure of overlap. It isn't a forecast of redundancy.
Here's an analogy. Your house is "exposed" to the weather. That doesn't mean it'll fall down. It might mean you need a better roof, or that you get a lovely sunny conservatory. Exposure tells you where change is likely to land. It doesn't tell you whether that change helps you or hurts you.
So why isn't exposure the same as losing your job? Because AI might take over some of your tasks, speed up others, or leave you free for work it can't do. Same exposure, very different outcomes.
The IMF: Big Numbers, With a Twist
In January 2024 the International Monetary Fund published a staff discussion note called Gen-AI: Artificial Intelligence and the Future of Work.
Its headline figure was that about 40% of employment worldwide is exposed to AI. In advanced economies like the UK and US it's about 60%. In emerging markets it's about 40%, and in low-income countries about 26%.
Now here's the twist most headlines skipped. The IMF split exposed jobs into two groups. In roughly half of those advanced-economy jobs, AI is likely to complement the work and help people be more productive. In the other half, AI could take over tasks, which might push down demand for that work or hold down wages.
So even the IMF's own "60%" isn't 60% of people losing their jobs. It's 60% where AI matters, and the effect could go either way.
The OECD: Automation Risk by Skills
The Organisation for Economic Co-operation and Development put out its Employment Outlook in July 2023.
It found that occupations at the highest risk of automation make up about 27% of employment on average across OECD countries. How did it decide what counts as high risk? It looked at jobs that use more than 25 of 100 skills and abilities that AI experts judged easy to automate.
The OECD also said there was little sign yet that AI was having a big effect on overall employment. It added that this could be because the change was still at an early stage. Notice how honest that is: it's a measure of risk, not of what had already happened.
The WEF: Asking Employers
The World Economic Forum's Future of Jobs Report 2025 came out in January 2025. Its method is different again. It's a survey of more than 1,000 large employers, covering over 14 million workers.
Those employers expected around 170 million jobs created and 92 million displaced by 2030. That's a net gain of about 78 million. And AI is only one of the drivers. The report also covers things like demographics and economic pressures.
So this is a set of employer expectations. Useful, yes. But it's what bosses think will happen, not something that has been measured.
The ONS: What UK Businesses Are Doing Now
The UK's Office for National Statistics does something different again. It doesn't estimate exposure at all. It measures adoption, meaning whether businesses say they're actually using AI.
Its article published on 20 July 2026, based on the Business Insights and Conditions Survey, found that around 35% of UK businesses with 10 or more employees reported using at least one AI technology in June 2026. That's up from around 12% in late 2023. Use varied hugely by sector: 58% in information and communication, but 13% in construction.
The ONS is upfront about the limits, too. The figures are self-reported, and they don't separate light use from AI being built deep into how a business works.
A Bonus: The ILO
One more worth knowing. In May 2025 the International Labour Organization and Poland's NASK published a refined global index. It found about one in four jobs worldwide potentially exposed to generative AI, with clerical jobs the most exposed. Their conclusion? Transformation, not replacement, is the most likely outcome for most jobs.
Your Three-Question Habit
So next time a statistic lands in your feed, run it through three quick questions:
- Who published it? A body like the IMF or ONS, a consultancy, or a vendor?
- When? AI moves fast, so a 2023 figure may already be dated.
- What does it measure? Exposure, automation risk, employer expectations, or actual adoption?
You'll find the answers in the report's summary or methods section, and usually in the first few pages.
Your Exercise
Now it's your turn. Pick two of the studies above, or another one you find, that mention your field or a field close to it. For each, write three or four sentences covering who published it, the date, what it measured, and what it says about work like yours.
Keep it short. Then add one line saying how you feel about it now that you know what it actually counts.
Key Takeaways
- "Exposure" means your tasks overlap with what AI can do. It isn't a prediction that you'll lose your job.
- IMF (January 2024): about 40% of global jobs exposed and about 60% in advanced economies, with around half of those likely to be complemented by AI.
- OECD (July 2023): about 27% of jobs in occupations at highest risk of automation.
- WEF (January 2025): an employer survey expecting 170m jobs created and 92m displaced by 2030.
- ONS (July 2026): about 35% of UK firms with 10+ staff reported using AI. That's adoption, not exposure.
- Always ask who published a figure, when, and what it measures.