Not all occupations have equal levels of risk related to AI. To provide a broader context, I started by looking at the average exposure by occupational group. The ILO assessed over 400 different occupations with respect to the amount of their work that could conceivably be completed by generative AI. On the average, these scores clustered around one major occupational category: clerical support. This type of work consists of text, data entry, record keeping, and general administrative duties. In fact, most of the same types of things that AI excels at doing nowadays.
Source: Author's analysis of ILO generative AI occupational exposure data (International Labour Organization, 2025).
Although AI is being adopted by many organisations, this does not mean every job will disappear. The impact depends on how much of a job’s tasks can be performed by AI. If most tasks are routine and predictable, the job has a higher risk of replacement. However, if only some tasks can be automated, the job is more likely to change rather than disappear. The chart below shows job exposure to AI and task variability across occupations. Jobs in the lower-right quadrant are the most vulnerable because they have high AI exposure and low task variability.
Source: Author's analysis of ILO generative AI occupational exposure data (International Labour Organization, 2025).
This graph helps answer the question posed at the beginning. The amount of exposure helps tell us what AI could potentially be able to accomplish, but companies will only automate if there is a considerable saving associated with it. To illustrate this fact, I combined ILO exposure data with US wage and job data. Each bubble on the graph represents one occupation based on its approximate size in terms of employee headcount. The data shows a clear pattern of low-wage, highly exposed occupations occupying much of it, and then there are high-wage and highly exposed occupations. This creates a ‘business reason’ to use AI/automation.
Source: Author's analysis of ILO generative AI occupational exposure data, ISCO-SOC occupation crosswalk data, and BLS Occupational Employment and Wage Statistics (International Labour Organization, 2025; U.S. Bureau of Labor Statistics, 2025).
Charts 1,2,3 show the link between AI exposure and economics, but employer action is also important. The World Economic Forum (2025) survey shows that employers are already responding to AI exposure in their workforce. The fastest-growing roles are expected to be mainly technology related, such as big data specialists, AI engineers and software developers. In contrast, the roles expected to decline fastest are administrative and clerical positions, which also had high AI exposure in the earlier charts. The color on each bar shows the exposure level, making the pattern clear: declining roles tend to have higher AI exposure.
Source: Growth and decline estimates derived from World Economic Forum Future of Jobs Report 2025, Figure 2.2; AI exposure scores from ILO generative AI occupational exposure data (World Economic Forum, 2025; International Labour Organization, 2025).
The final chart looks at scale: how many dollars are actually tied to these positions exposed to AI? I determined this total by taking the number of employees in these occupations in the US and multiplying it by the average annual salary of these employees. This gives the total amount of money paid to these employees; that is, the total amount of money being paid by companies to employ people doing these jobs. Looking only at software developers, they account for approximately $250 billion in wages. The next largest is the combination of office clerks and secretaries, equal to about $200 billion in wages for all of those positions. These numbers represent the attractiveness of potential AI investments. However, replacing millions of employees requires major infrastructure, system integration and ongoing costs that are not directly reflected in basic comparisons.
Source: Author's analysis of ILO generative AI occupational exposure data, ISCO-SOC occupation crosswalk data, and BLS Occupational Employment and Wage Statistics (International Labour Organization, 2025; U.S. Bureau of Labor Statistics, 2025).
The AI threat to jobs is not uniform across the board: it is concentrated in clerical and administrative work, though even here, replacement makes sense only when the economics are right. The greater risk is perhaps most true in higher incomes, where companies stand to save more per worker.
For most workers, AI is more likely to reshape their role than remove it. But with trillions in wages sitting in exposed occupations, the financial incentive to automate is real and growing. We are now talking about not if AI can do the job. We are talking about whether it is worth the cost.
I used ChatGPT (OpenAI, 2026) and Claude (Anthropic, 2026) as most of learning support tools during this assignment. These tools were used to assist with advise on story angle, improving the clarity of written explanations, debugging R coding errors, and checking whether the visual narrative was clearly structured.
The final topic selection, data source selection, data analysis decisions, chart design choices, interpretation of results, and final written content were reviewed, tested, edited, and approved by me before submission.
No artificial data was used. All visualisations were created in R studio using publicly accessible data sources, and all sources used in the assignment have been acknowledged and referenced.
I also referred to publicly available YouTube learning resources, including Riffomonas Project, to better understand R-based data visualisation practices using ggplot2. These resources were used for general learning and inspiration only. The final code, chart structure, data analysis decisions, and written interpretation were developed, tested, and edited by me for this assignment.
Gmyrek, P., Berg, J., Kamiński, K., Konopczyński, F., Ładna, A., Nafradi, B., Rosłaniec, K., & Troszyński, M. (2025). Generative AI and jobs: A refined global index of occupational exposure (ILO Working Paper No. 140). International Labour Organization. https://doi.org/10.54394/HETP0387
Ritchie, H., Roser, M., & Rosado, P. (2025). Data centres and electricity demand. Our World in Data. https://ourworldindata.org/grapher/data-centers-share-electricity-demand
U.S. Bureau of Labor Statistics. (2025). Occupational employment and wage statistics: May 2025 national estimates. U.S. Department of Labor. https://www.bls.gov/oes/tables.htm
World Economic Forum. (2025). The future of jobs report 2025. World Economic Forum. https://www.weforum.org/publications/the-future-of-jobs-report-2025/
Anthropic. (2026). Claude [Large language model]. https://claude.ai/
OpenAI. (2026). ChatGPT [Large language model]. https://chat.openai.com/
Riffomonas Project. (n.d.). YouTube. Retrieved from https://www.youtube.com/@Riffomonas