📈 Chartpack: Generative AI and white-collar jobs

What the emerging academic literature tells us

ChatGPT was released a mere six months ago, but it has already seen massive adoption around the world, reaching 100 million users in only 2 months and kicking off the era of generative AI (GenAI) [read our chartpack on the state of GenAI here]. This adoption has quickly spread to the workplace. A survey done by the professional networking app Fishbowl in January 2023 found that already 43% of respondents were using ChatGPT in their workplace.

Following this rapid adoption of generative AI such as ChatGPT, we’re reviewing preliminary academic research in today’s Chartpack to help guide strategic decision-making around GenAI integration. Please note that these studies focus on direct impacts on work tasks. Second-order effects such as potential wage impacts or job creation will be addressed in future Chartpacks for premium members.

Who will be exposed to GenAI?

A study published in March 2023 by Eloundou et al. examined the potential effects of GPTs¹ on the labour market, specifically focusing on the tasks they could complete or augment within each profession. They found that 80% of the U.S. workforce could have at least 10% of their work tasks affected by LLMs². This is over 130 million people in the U.S. alone. Moreover, this impact rises significantly with income, affecting tasks within white-collar jobs the most. 

The study predicts that jobs that require a higher level of education or training are most likely to be exposed to LLMs. Jobs requiring a bachelor’s degree (2-4 years of training) could be most affected, with over 30% of these jobs expected to have at least 50% of their tasks exposed to LLMs. Examples of highly exposed jobs include information services, finance, publishing and telecommunications.

How is GenAI affecting these occupations? 

While the academic literature is still in its infancy, a few initial studies on writing, coding and call-centre performance provide indications of GenAI’s initial effects.

All of these studies found an increase in performance. Let us explain how.