Microsoft Study: Jobs Least & Most Impacted by Generative AI

Microsoft study names 20 jobs least likely to be changed by generative AI — healthcare and blue‑collar work top the list

A Microsoft study using about 200,000 anonymized Bing Copilot conversations finds hands-on healthcare and blue‑collar occupations have the lowest AI applicability today, while communication-heavy knowledge work shows the greatest exposure to text-based generative AI tools.

Key takeaways

• Hands-on roles — bedside care, emergency response, and manual trades show the lowest overlap with current text-based generative AI.
• Applicability ≠ displacement — the study measures task overlap with AI capabilities, not job loss or unemployment.
• High-exposure work — language, document, and routine information tasks (writing, translating, customer service) are most supported by generative AI.
• Local planning implications — regions like Sonoma, California can use the findings to prioritize vocational training and upskilling.

How the Microsoft study was done

Microsoft researchers analyzed roughly 200,000 anonymized Bing Copilot conversations from 2024 to identify what people asked text-based generative AI to do. They mapped those tasks to U.S. occupational taxonomies (O*NET and Bureau of Labor Statistics) and computed an “AI applicability score” that measures task-level overlap between present-day generative AI capabilities and job activities.

For full methodology, see the Microsoft research summary Working with AI: Measuring the Applicability of Generative AI to Occupations and the companion arXiv paper.

Jobs least likely to be changed by generative AI

The study and subsequent coverage highlight a consistent pattern: occupations that require physical presence, fine motor skills, in-person care, or work in unpredictable environments show the lowest AI applicability. Below are 20 examples identified as among the least likely to be affected by current text-based generative AI tools.

1. Roofers (and helpers) — outdoor, physical construction work
2. Tire builders — hands-on industrial assembly
3. Gas compressor and pumping station operators — manual operation of field equipment
4. Industrial truck and tractor operators — forklift and heavy vehicle work
5. Cement masons and concrete finishers — hands-on material work
6. Machine feeders and offbearers — loading/unloading for manufacturing
7. Packaging and filling machine operators — physical production tasks
8. Dishwashers — manual kitchen work
9. Surgical assistants — intra‑operative, hands‑on support
10. Massage therapists — close-contact therapeutic work
11. Ophthalmic medical technicians — patient testing and equipment use in eye care
12. Medical equipment preparers — sterilizing and readying instruments
13. First-line supervisors of firefighting and prevention workers — field coordination under risk
14. Firefighters — emergency response with unpredictable conditions
15. Construction laborers — varied on-site physical labor
16. Home health and personal care aides — direct patient support and care
17. Nursing assistants and orderlies — bedside patient handling and movement
18. Plumbers, pipefitters, and steamfitters — in-person diagnostics and repair
19. Electricians — safety-critical on-site installation and troubleshooting
20. Truck and heavy vehicle roles — lower applicability for text-based AI though robotics/autonomy could change this

Jobs most exposed to generative AI

By contrast, occupations centered on language, documents, and routine information tasks show the highest AI applicability. Microsoft and media coverage point to roles such as interpreters and translators, writers and authors, customer service representatives, technical writers, editors, telemarketers, and many office and sales positions as already benefiting from generative AI for drafting, summarizing, translating, and answering queries.

See coverage summarizing high-exposure occupations at GeekWire and Fortune.

What Microsoft says about applicability vs. displacement

“Applicability ≠ displacement.”

Microsoft stresses that the AI applicability score measures current task support by text-based generative AI, not whether entire occupations will vanish. People using AI to augment work may see productivity gains and evolving job tasks; the study aims to identify where current AI tools can help today, not to forecast exact job losses. For Microsoft’s discussion, read the research blog on applicability vs. displacement.

Caveats and limits of the research

• Task-level measure: the score reflects overlap at the task level, not whole-occupation elimination — see the Microsoft research page for details.
• Text-only focus: analysis centers on text-based generative AI; robotics, sensors, computer vision, and autonomy could affect other job categories differently.
• Adoption matters: labor-market outcomes depend on firm adoption, regulation, and worker retraining; see Microsoft WorkLab on AI at work for policy context.

Relevant reads: Microsoft research paper, the arXiv version, and the Microsoft WorkLab piece on AI at work.

Implications for Sonoma, California

Economic impact: Sonoma’s economy — agriculture, tourism, construction, healthcare, and small business services — includes many of the low-applicability roles identified by Microsoft (roofers, construction laborers, plumbers, electricians, nursing assistants). These jobs may face less direct disruption from generative AI in the near term, suggesting local demand could remain steady or grow as housing and care needs expand.

Healthcare and caregiving: Sonoma’s hospitals, clinics, and in‑home care providers rely on bedside nurses, nursing assistants, and home health aides. The Microsoft findings imply these hands-on roles will remain centered on human care for now — a key point for a county with an aging population.

Blue-collar workforce and training: Trades such as plumbing, electrical work, roofing, and concrete finishing are likely to remain in demand. Local community colleges and trade schools can strengthen training and apprenticeship programs to align with employer needs.

Office and communication jobs: Wineries, tourism, and small firms often depend on marketing, PR, customer service, and office staff — areas with higher AI applicability. Employers may adopt generative AI for tasks like drafting copy, responding to customer inquiries, or translating materials; worker upskilling in AI tools will be useful.

Public safety: Firefighters and first-line supervisors retain low AI applicability due to unpredictable, physical emergency work. In wildfire-prone Sonoma County, human skills in firefighting and emergency coordination remain essential, even if AI supports planning and mapping.

Policy and planning: Local leaders should use this research as a planning tool: support vocational training, caregiving education, and AI upskilling for office workers. Consider grants or programs that help small businesses adopt AI safely and equitably.

Sources and further reading

• Microsoft research paper — Working with AI: Measuring the Applicability of Generative AI to Occupations
• arXiv PDF of the Microsoft paper
• Microsoft research blog on applicability vs. displacement
• GeekWire coverage: top 25 least-impacted jobs
• Fortune summary of Microsoft research and occupational impact
• Microsoft WorkLab — AI at work

Reporting in this article draws on Microsoft’s public research and major coverage to explain how generative AI’s occupational impact varies across jobs and what that means for local economies like Sonoma.