Key Points:
- The share of job postings mentioning GenAI or related terms has skyrocketed over the past year, up 3.5 times in the United States and even faster in other countries.
- Data analytics consistently ranks as the sector with the highest GenAI postings share across all nine analyzed countries. Scientific research and software development also rank highly in eight of these countries. However, beyond these three sectors, GenAI usage varies considerably by country.
- Global trends show consistent patterns in sectors where GenAI-related job postings exceed expectations (scientific research, arts & entertainment, industrial engineering) and those where usage falls short (medical information, insurance, logistic support).
- In the United States, underperformers were consistent with global trends, with GenAI-related job postings below expectations in the insurance, logistic support, and medical information sectors. The US sectors that exceeded expectations, including architecture, arts and entertainment, and industrial engineering, were also consistent with global trends.
Over the past year, job postings mentioning generative artificial intelligence (GenAI) or related phrases have increased dramatically. In the United States, the share of job postings referencing GenAI more than tripled between September 2023 and September 2024, and growth has been even faster in other nations. A deeper analysis shows that the sectors in which GenAI adoption is highest in the US (including data analytics and software development) are also largely the same sectors where the current generation of GenAI tools is more likely to replace a human at a higher share of common skills. But there are also some surprising sectors where GenAI adoption is underperforming or overperforming expectations, in the US and worldwide.
Does GenAI usage align with expectations?
This analysis builds on previous Hiring Lab work that assessed the likelihood of current-generation GenAI tools replacing a human at any one of more than 2,800 individual work skills on a scale ranging from “very unlikely” to “unlikely,” “potentially,” “likely,” and “very likely.” Sectors with the highest replacement likelihood — defined as sectors where the replacement likelihood for job-relevant skills was most commonly rated “possible,” “likely,” or “very likely” — were typically found in knowledge-intensive jobs, particularly those involving repetitive or data-driven tasks. The tech sector and various back-office roles, including accounting, marketing, administrative assistance, and human resources, were among those with a high theoretical exposure to GenAI.
This work explores whether workplace usage of GenAI aligns with our previous assessment of GenAI’s ability to replace specific skills. If so, we’d expect the share of GenAI-related job postings to be highest in occupations with the highest share of skills “likely” or “potentially” replaced by GenAI, and minimal in jobs with a lower replacement likelihood. We conducted this analysis across nine countries using September 2024 data.
We found that sectors where GenAI is most likely to replace certain skills typically have a higher share of job postings mentioning these tools. But there are some surprises: Some sectors with a relatively high share of skills likely to be replaced (including accounting and insurance) have virtually no GenAI-related job postings. At the same time, GenAI is mentioned more often than expected in some sectors with a low replacement likelihood (like scientific research and arts & entertainment). In all countries analyzed except the US, GenAI references are typically concentrated in a small number of sectors, with many sectors — including both high- and low-replacement likelihood sectors — having no job postings that mention GenAI.
Some deviation from expectations is to be expected. Hiring Lab’s earlier analysis only tested the capabilities of OpenAI’s GPT-4o model and did not assess any other GenAI tools — some of which are highly specialized towards specific skills. And while GenAI may be able to perform a range of skills at a high level, workplace adoption may still lag behind due to a lack of infrastructure or processes to facilitate its usage, along with regulatory limitations or even moral or cultural concerns.
Where is GenAI most commonly found?
Across all nine countries analyzed — Australia, Canada, France, Germany, Ireland, Singapore, Spain, the United Kingdom, and the United States — data analytics was the sector where GenAI was most likely to be mentioned in job descriptions. Roles within data analytics, including data scientists and/or data engineers, often work either directly developing GenAI tools, or heavily incorporating GenAI into their workflows. In the US, 5.1% of data analytics job postings referenced GenAI in September, a middle-of-the-pack reading that was higher than the shares in the UK, Germany, and France, but lower than shares in Australia, Canada, Ireland, Singapore, and Spain.
GenAI is also prominent in software development and scientific research postings, ranking among the top five sectors for GenAI postings in eight of the nine countries analyzed. Software development missed out in Germany, and scientific research in Ireland.
Beyond these sectors, references to GenAI tools vary more from country to country. In Canada, Ireland, and Spain, marketing job postings often mention GenAI, while media & communications stand out in the US and the UK. In Singapore, unlike most other countries analyzed, GenAI is frequently mentioned in medical information and architecture roles.
These variations underscore the lack of a universal roadmap for GenAI adoption. Adoption rates differ by country and sector, leading to unique trends outside the few tech-related sectors where GenAI is central to performing the role. Even within those tech sectors, like data analytics, references to GenAI within job descriptions can vary significantly across regions.
GenAI trends are broadly consistent with expectations
Growth in GenAI-related job postings aligns broadly with expectations based on each occupation’s potential GenAI replacement likelihood. However, there were still a few surprises.
Observations from data in September 2024:
- A positive relationship exists between the share of GenAI postings and GenAI replacement likelihood. The linear correlation coefficient is 0.52 in the US, and ranges from 0.42 in Australia to 0.57 in France, indicating that sectors with higher exposure to GenAI often have a larger share of job postings mentioning GenAI tools.
- GenAI usage is most widespread across sectors in the United States. In September, only four occupational categories (pharmacy, childcare, personal care & home health, and beauty & wellness) in the United States had no GenAI mentions, compared to 22 sectors in Australia, 13 sectors in Canada and 10 sectors in the UK.
- Some sectors with a high-replacement likelihood, including accounting in Australia and insurance or medical information in other nations, had no GenAI-related job postings in September 2024. That does not necessarily mean that workers in these industries aren’t using these tools, but it suggests that GenAI-based skills are not crucial for recruitment.
- Some sectors with only moderate theoretical exposure (where GenAI is ‘likely’ or ‘possible’ to replace fewer than 55% of skills), including the scientific research and arts & entertainment sectors, have a relatively high number of GenAI-related job postings.
Unexpected trends in GenAI adoption
To identify the biggest surprises in GenAI adoption, we established two criteria. These criteria identify occupations where GenAI is mentioned more often or less often than expected, given the sector’s skill-replacement likelihood and each nation’s GenAI posting share.
- Fewer GenAI postings than expected: If a sector’s posting share was less than one-third the national average and the skill replacement likelihood was 55% or above.
- More GenAI postings than expected: If a sector’s posting share was more than 1.5 times the national average and the skill replacement likelihood was less than 55%, or if the sector’s posting share was greater than the national average and the skill replacement likelihood was below 45%.
Several patterns emerged across regions. The GenAI posting share was lower than expected in sectors including medical information (eight countries), insurance (seven countries), and logistic support (four countries). It was also relatively low in accounting and administrative assistance (three countries each). In the United States, underperformers were consistent with global trends, with GenAI-related job postings below expectations in the insurance, logistic support, and medical information sectors.
In these sectors, GenAI may be able to perform a range of skills at a high level. However, workplace adoption may lag behind due to various factors, such as a lack of infrastructure or processes to facilitate its usage, regulatory limitations, or even moral or cultural concerns. Some of these factors may be addressed in the near term, and we certainly expect businesses globally to invest heavily in their AI capacity. Still, others may prove more difficult to overcome. The regulatory response from national or regional governments will play a key role in how GenAI usage evolves.
According to the International Monetary Fund (IMF), some occupations are shielded from AI-driven job displacement. That level of protection or shielding varies across countries and will change over time. Occupations with a high level of shielding, including judges and medical practitioners, where decisions have significant consequences and where mistakes can be more costly, are less likely to experience displacement due to GenAI technologies, even if the tools themselves can perform those skills.
Conversely, GenAI adoption exceeded expectations in a range of sectors. Scientific research was the most obvious example, exceeding expectations in eight countries (all bar Ireland), followed by architecture (six countries), industrial engineering (six countries), and arts & entertainment (five countries). The sectors that exceeded expectations in the United States, including architecture, arts & entertainment, and industrial engineering, were also consistent with global trends.
The higher-than-expected GenAI uptake in some sectors may reflect the emergence of new, highly specialized GenAI tools. Indeed’s analysis focused on ChatGPT’s capabilities, but there are now tools specifically designed for tasks like image creation or video processing, which may explain the higher-than-expected GenAI postings in sectors like arts & entertainment, or even architecture.
Assessment
Our research identifies two key takeaways. First, GenAI’s rapid rise has primarily been driven by sectors where skills have a relatively high-replacement likelihood, but there were notable exceptions. Second, GenAI job postings are heavily concentrated in a relatively small number of sectors, suggesting that while GenAI technology is advancing, its practical value remains unproven in many sectors — even those with high exposure.
Methodology
The analysis involved extracting job postings directly related to Generative AI, using specific keywords indicating its presence, such as ‘Generative AI,’ ‘Large Language Models,’ and ‘ChatGPT.’ For our methodology on country replacement likelihood estimates, including the most- and least-exposed occupations, please see our post here.
The graph below shows the occupations with the most and least exposure to GenAI.
