Handa, K., Tamkin, A., McCain, M., Huang, S., Durmus, E., Heck, S., Mueller, J., Hong, J., Ritchie, S., Belonax, T., Troy, K. K., Amodei, D., Kaplan, J., Clark, J., & Ganguli, D. (n.d.). Which Economic Tasks are Performed with AI? Evidence from Millions of Claude Conversations.
High-level summary of the paper
The study by Handa et al. examines how artificial intelligence (AI), specifically through the AI assistant Claude, is being used across different economic sectors. By analysing over four million anonymised conversations, the authors map AI interactions to occupational tasks using the U.S. Department of Labor’s ONET database.
The findings reveal key insights into AI adoption patterns, showing that:
- AI usage is concentrated in software development and writing-related tasks, accounting for nearly 50% of all usage
- AI is being integrated into 36% of occupations, where at least a quarter of job-related tasks involve AI
- High-wage jobs (such as those requiring a bachelor’s degree) see the highest AI adoption, while both low-wage and very high-wage jobs (such as restaurant workers and physicians) show lower adoption
- AI is used both for augmentation (57%) (collaborative, iterative tasks) and automation (43%) (fully executing tasks without human involvement)
- Jobs that require cognitive skills (e.g., critical thinking, reading comprehension, and writing) show high AI engagement, while those requiring physical skills (e.g., equipment maintenance, manual labour) show minimal interaction
Key takeaways for university leadership
For leaders in higher education, this paper underscores the urgency of reassessing curriculum design, digital strategy, and workforce preparedness in light of AI’s growing role in professional tasks. Here are the key implications:
- AI is already reshaping the skills economy
- AI is not just a futuristic concept; it is already being used to perform tasks that were previously considered human-driven
- Many professional occupations (writing, software development, business analytics) are rapidly integrating AI, suggesting that AI literacy is now as fundamental as digital literacy
- Higher education must shift towards AI-augmented learning
- Traditional teaching models that focus on content recall are becoming obsolete. Instead, students need to learn how to work alongside AI—leveraging it for research, writing, data analysis, and problem-solving
- Courses should integrate critical thinking and AI auditing skills—ensuring that students know when to trust AI outputs and how to refine them
- The automation-augmentation balance is crucial
- The study indicates that AI is being used both for augmentation (helping users think, iterate, and improve) and automation (fully executing tasks)
- Universities need to prepare students for both: how to leverage AI as a tool and how to maintain human oversight over automated processes
- AI exposure is highly unequal across disciplines
- STEM fields, particularly computer science and data analytics, are at the forefront of AI usage
- However, creative fields like writing, media, and business also see heavy AI adoption
- Health and physical sciences remain relatively untouched, but this will likely change as AI capabilities expand
- University leaders must ensure that AI integration is equitable across all disciplines, not just tech-heavy courses
- AI is penetrating the workforce at different depths
- In some jobs, AI is used sporadically (e.g., brainstorming ideas, summarising content), while in others, it is deeply embedded (e.g., coding, content generation)
- This means graduates will need different levels of AI competency depending on their field, requiring tailored AI literacy programmes
Practical implications: What you can do tomorrow
If you are in a leadership role at a university, here are immediate steps you can take to act on these findings:
1. Introduce an AI literacy programme for staff and students
- Host workshops on AI ethics, usage, and limitations
- Offer AI skill-building modules that teach students how to use AI effectively and responsibly in their respective disciplines
- Develop a mandatory AI awareness course similar to digital literacy requirements
2. Revise curriculum and assessment strategies
- Shift assessments from rote learning to tasks that require AI-assisted critical thinking
- Design coursework that encourages iterative AI collaboration (e.g., students draft content, use AI for feedback, then refine their work)
- Create AI-integrated case studies for business, healthcare, law, and humanities students
3. Encourage AI-assisted research and productivity
- Provide faculty training on how to integrate AI into their research
- Encourage staff to explore AI for automating administrative tasks (e.g., drafting reports, summarising research)
- Implement AI for institutional decision-making—such as analysing student performance trends and workload distribution
4. Establish an AI governance framework
- Develop clear AI usage policies for students and staff (e.g., when it is acceptable to use AI in assignments, research, and administrative work)
- Create an AI ethics committee to regularly assess AI’s impact on learning and teaching
- Ensure transparency in AI usage—educators should disclose when AI-generated content is used in learning materials
5. Build AI partnerships with industry
- Collaborate with tech firms and AI developers to integrate real-world AI applications into student projects
- Work with local businesses to identify AI-related skills that employers expect from graduates
- Develop AI-focused internships where students gain hands-on experience with AI-driven workflows
Final thoughts
This paper confirms what many already suspect—AI is rapidly becoming embedded in the professional world. Universities that fail to adapt will risk producing graduates who are ill-prepared for a workforce increasingly reliant on AI. By embedding AI into curricula, upskilling staff, and fostering AI-conscious leadership, universities can ensure they are not just reacting to change but shaping the future of AI-powered education.