From the Founder's Desk: Microsoft Research Reveals Which Occupations Are Using AI - and How Successfully
By: Chris Engle, CEO & Founder, Headlight Data
Recent research from Microsoft provides comprehensive, real-world evidence of how Americans are actually using generative AI at work — and which occupations are finding it most useful.
Rather than predicting what AI might do to jobs in the future, Microsoft analyzed 200,000 real conversations from 2024 between users and Bing Copilot (which runs on OpenAI’s large language models, the same technology as ChatGPT) to see what work activities people are already seeking AI assistance with, and how well the AI performs.
Microsoft created an “AI applicability score” for 785 occupations covering 150 million U.S. workers that combined three measures: how often AI is used today, how successfully, and for how much of each work activity.
Knowledge work and communication roles show the highest AI applicability:
- Information gathering, writing, and editing
- Customer service and sales communication
- Research and data analysis
- Technical documentation
Physical and manual work shows minimal current applicability — not because these jobs are “safe” from technology forever, but because today's large language models simply don't interface with physical tasks.
Highest AI Applicability (occupations where AI is being used most successfully):
- Interpreters and Translators (top score: 0.49/1.0)
- Writers and Authors (0.45)
- Customer Service Representatives (0.44) — affecting 2.9 million workers
- Sales Representatives (0.46) — affecting 1.1 million workers
- Data Scientists and Analysts (0.32-0.35)
Lowest AI Applicability:
- Construction and manual labor occupations
- Heavy equipment operators
- Healthcare support roles requiring physical presence
- Food service and cleaning occupations
Today’s “AI Applicability” were highest for both high-skill fields in Science/Computer and low-skill fields in Office Support, as shown below:
My key takeaway from this research is that AI is currently better at assisting than replacing jobs. As AI capabilities evolve and new tools emerge, we’ll continue to monitor research on how these technologies actually get used in the workplace — and how “assisting” workers may shift into “replacing” workers, or at least limiting new hires due to AI as many CEOs have suggested they plan to do.
As you evaluate AI's impact in your community, keep observing:
- Which local industries and workers are already finding AI useful?
- What’s the gap between potential and adoption?
- How do we measure or observe productivity gains at the worker-level?
- How can we ensure that AI creates a winning scenario for our community, industries, and workers?
For more details: The full research paper and data are available for PDF download.