For most jobs, AI changes the tasks long before it threatens the role.
The office manager of a small property management company reads yet another headline predicting that AI will wipe out millions of jobs. She looks at her own role: scheduling maintenance, drafting letters to tenants, processing invoices, answering the same questions every day. A lot of it, she realises, an AI tool could already help with. She feels a knot of worry.
Her employer feels something different: uncertainty. Should he be planning to replace roles, retrain people, or simply wait and see?
The loudest headlines about AI and jobs tend to swing between utopia and catastrophe. One of the most careful recent studies, from the United Nations’ labour agency, offers a more measured picture. Its central message is that AI is far more likely to change what people do at work than to make them unnecessary, and that how businesses manage that change will matter enormously.
For most jobs, AI changes the tasks long before it threatens the role.
What the ILO found
In May 2025, the International Labour Organization and Poland’s National Research Institute (NASK) published Generative AI and Jobs: A Refined Global Index of Occupational Exposure. The researchers analysed nearly 30,000 job tasks, combined with worker surveys, expert reviews and AI-assisted scoring, to estimate how exposed different occupations are to generative AI.
They found that around one in four jobs worldwide is in an occupation potentially exposed to generative AI. Exposure was much higher in high-income countries, at about 34% of employment, than in low-income countries, at about 11%. Clerical and administrative roles were the most exposed, followed by some highly digital jobs in media, software and finance.
Crucially, only a small share of jobs, around 3.3%, fell into the highest exposure category, where most tasks could feasibly be done by AI without human input. Even there, almost no jobs consisted entirely of automatable tasks. The authors concluded that transformation of jobs is the most likely impact of generative AI.
The study also found that women were more exposed than men, largely because women are overrepresented in clerical roles. And it stressed that the figures reflect potential exposure, not actual job losses. Real-world effects depend on infrastructure, skills, costs and choices made by employers and governments.
For South Africa, an upper-middle-income country with a large clerical and service workforce, the findings suggest significant change ahead, but not an inevitable wave of job losses.
Very few jobs consist entirely of tasks AI could do alone.
Preparing roles for transformation
For small business owners, the practical question is how to help roles evolve, rather than whether to eliminate them. A few steps help.
Look at tasks, not job titles. Break each role into its main tasks. Some will be well suited to AI assistance, such as drafting standard letters, summarising information or scheduling. Others rely on judgement, relationships and local knowledge that AI can’t replace.
Redirect time to higher-value work. When AI takes over routine tasks, decide deliberately what people will do with the time freed up: more time with customers, better follow-up, problem-solving, quality checks or new services.
Invest in skills. As explored in Nearly Two in Five Work Skills Will Change by 2030, learning to work alongside AI, and strengthening the human skills it can’t replicate, will matter for almost every role.
Involve people in the change. The ILO urged employers and workers to work through these transitions together. In a small business, that simply means talking openly with staff about how their work might change, and asking for their ideas.
Pay attention to who is most affected. Administrative roles, often held by women, are among the most exposed. Thoughtful planning can help these team members move into more varied, valuable roles rather than being left behind.
INFOGRAPHIC GOES HERE
Exposure to generative AI. A simple bar chart: Share of jobs in exposed occupations — high-income countries: 34%; world: about 25%; low-income countries: 11%. 3% of jobs. Caption: ILO–NASK (2025): transformation of jobs is the most likely impact.
Look at tasks → Redirect time → Grow the role.
A role-by-role conversation
Pick one role in your business, perhaps the most administrative one. Sit down with the person in it and list their main tasks together. Mark which ones AI could help with, which ones it couldn’t, and which ones they’d love more time for. Then agree on one experiment: an AI tool for one routine task, and a plan for using the time it frees up. Review it after a month.
A few questions can help:
- Which tasks in each role could AI assist with, and which depend on human judgement?
- What valuable work would we do more of if routine tasks took less time?
- What skills will our team need to work well alongside AI?
- Have we talked openly with staff about how AI might change their work?
Which brings us back to the property management office manager and her knot of worry. The research suggests her job is far more likely to change than to disappear. AI could draft the routine letters and sort the invoices, freeing her to spend more time resolving tenant issues, coordinating contractors and keeping owners informed, the parts of her job that depend on knowing people and properties. With a thoughtful employer, that change could make her role more valuable, not less.
The question is how to help roles evolve, not whether to eliminate them.
One Like Research & Further Reading
- International Labour Organization and NASK, Generative AI and Jobs: A Refined Global Index of Occupational Exposure, ILO Working Paper, May 2025. Summarised by UN News, 20 May 2025.