For millions of workers, artificial intelligence has become the Grim Reaper stalking their careers, or so it might feel.
While many people are excited about the dawn of the AI age, others worry that the technology will soon replace them at work.
Recently, researchers at Microsoft looked at a slew of careers and issued a report that reveals the 40 jobs that AI threatens most — and the 40 jobs most likely to escape unscathed by the technology — at least for now.
Microsoft says AI is changing the world quickly. Almost 40% of Americans say they use generative AI at home or work. That means the technology is “outpacing the early diffusion of the personal computer and the internet,” according to the report.
As part of its analysis, researchers pored over a dataset of 200,000 anonymized conversations between users and Microsoft Copilot, which is a generative AI system available to the public.
The analysis revealed that workers most often turn to AI for help with gathering information and writing. AI itself most often provides information and assistance, writing, teaching and advising.
The most vulnerable jobs
Researchers looked at the extent to which AI is being used to complete work tasks in each occupation and how many of those tasks AI is able to complete successfully to arrive at an “AI applicability score” for each job. According to Microsoft:
“This score captures if there is nontrivial AI usage that successfully completes activities corresponding to significant portions of an occupation’s tasks.”
In short, the higher a job’s AI applicability score, the more likely it is to be impacted by AI.
Microsoft found that, in general, the highest AI applicability scores went to jobs involving knowledge work (such as those in the computer, mathematical, and office and administrative support fields) or providing and communicating information (such as sales jobs).
The specific jobs with the highest AI applicability scores are:
- Interpreters and translators: AI applicability score of 49%; current employment level of 51,560 jobs
- Historians: 48%; 3,040
- Passenger attendants: 47%; 20,190
- Sales representatives of services: 46%; 1,142,020
- Writers and authors: 45%; 49,450
- Customer service representatives: 44%; 2,858,710
- CNC tool programmers: 44%; 28,030
- Telephone operators: 42%; 4,600
- Ticket agents and travel clerks: 41%; 119,270
- Broadcast announcers and radio DJs: 41%; 25,070
- Brokerage clerks: 41%; 48,060
- Farm and home management educators: 41%; 8,110
- Telemarketers: 40%; 81,580
- Concierges: 40%; 41,020
- Political scientists: 39%; 5,580
- News analysts, reporters and journalists: 39%; 45,020
- Mathematicians: 39%; 2,220
- Technical writers: 38%; 47,970
- Proofreaders and copy markers: 38%; 5,490
- Hosts and hostesses: 37%; 425,020
- Editors: 37%; 95,700
- Postsecondary business teachers: 37%; 82,980
- Public relations specialists: 36%; 275,550
- Demonstrators and product promoters: 36%; 50,790
- Advertising sales agents: 36%; 108,100
- New accounts clerks: 36%; 41,180
- Statistical assistants: 36%; 7,200
- Counter and rental clerks: 36%; 390,300
- Data scientists: 36%; 192,710
- Personal financial advisors: 35%; 272,190
- Archivists: 35%; 7,150
- Postsecondary economics teachers: 35%; 12,210
- Web developers: 35%; 85,350
- Management analysts: 35%; 838,140
- Geographers: 35%; 1,460
- Models: 35%; 3,090
- Market research analysts: 35%; 846,370
- Public safety telecommunicators: 35%; 97,820
- Switchboard operators: 35%; 43,830
- Postsecondary library science teachers: 34%; 4,220
The least vulnerable jobs
Before you panic that AI might soon kill your career, it’s important to note that Microsoft says its analysis does not suggest that AI is performing all the work activities of any single occupation. The report states:
“…the overlap between AI capabilities and various occupations is very uneven. There are definitely some occupations for which many — perhaps even most — work activities have some overlap with demonstrated AI capabilities. But even when there is overlap, the task completion rate is not 100% and the scope of impact is usually moderate.”
With that said, the jobs with the lowest AI applicability scores are:
- Phlebotomists: AI applicability score of 3%; current employment level of 137,080 jobs
- Nursing assistants: 3%; 1,351,760
- Hazardous materials removal workers: 3%; 49,960
- Helpers of painters, plasterers and stucco masons: 3%; 7,700
- Embalmers: 3%; 3,380
- Plant and system operators, all other: 3%; 15,370
- Oral and maxillofacial surgeons: 3%; 4,160
- Automotive glass installers and repairers: 3%; 16,890
- Ship engineers: 3%; 8,860
- Tire repairers and changers: 2%; 101,520
- Prosthodontists: 2%; 570
- Helpers of production workers: 2%; 181,810
- Highway maintenance workers: 2%; 150,860
- Medical equipment preparers: 2%; 66,790
- Packaging and filling machine operators: 2%; 371,600
- Machine feeders and offbearers: 2%; 44,500
- Dishwashers: 2%; 463,940
- Cement masons and concrete finishers: 1%; 203,560
- Supervisors of firefighters: 1%; 84,120
- Industrial truck and tractor operators: 1%; 778,920
- Ophthalmic medical technicians: 1%; 73,390
- Massage therapists: 1%; 92,650
- Surgical assistants: 1%; 18,780
- Tire builders: 1%; 20,660
- Helpers of roofers: 1%; 4,540
- Gas compressor and gas pumping station operators: 1%; 4,400
- Roofers: 1%; 135,140
- Oil and gas roustabouts: 1%; 43,830
- Maids and housekeeping cleaners: 1%; 836,230
- Paving, surfacing, and tamping equipment operators: 1%; 43,080
- Logging equipment operators: 1%; 23,720
- Motorboat operators: 0%; 2,710
- Orderlies: 0%; 48,710
- Floor sanders and finishers: 0%; 5,070
- Pile driver operators: 0%; 3,010
- Rail-track laying and maintenance equipment operators: 0%; 18,770
- Foundry mold and coremakers: 0%; 11,780
- Water treatment plant and system operators: 0%; 120,710
- Bridge and lock tenders: 0%; 3,460
- Dredge operators: 0%; 940
Microsoft researchers did not find a significant link between AI applicability scores and educational requirements for jobs. AI impacts occupations that require a bachelor’s degree only “slightly more” than jobs that require less education.
The same was true of wages: The average AI applicability score among high-paying jobs is only “slightly higher” than the average score among lower-paying jobs.
Add a Comment