Scoring 342 US jobs on AI/robot risk: office work scores 94%, hands-on care 17%, and the riskier jobs pay more
A new analysis scores 342 US jobs on AI/robot risk, revealing that office work faces a 94% risk, while hands-on care has a 17% risk. Interestingly, the study found that riskier jobs tend to pay more, with the 60 "Very High Risk" jobs having a median salary of $79,665, compared to $51,030 for the 27 lowest-risk jobs. This trend contrasts with previous automation waves that primarily impacted lower-paid routine work.
时间与来源
时间显示为 UTC
显示时区：UTC
本地时区尚不可用，暂时显示 UTC。
发布当时偏移：UTC+02026年10月29日 08:47 UTC
收录当时偏移：UTC+02026年10月29日 14:01 UTC
- 发布
- 2026年10月29日 08:47
- 收录
- 2026年10月29日 14:01
- 来源类型
- 开发者社区
- 档位
- 社区
- 信源状态
- 正常
档位是按信源手工设定的编辑判断，不是逐条打分。
讨论趋势
百分比基于采集到的讨论信号，不代表新增评论数或独立参与人数。曲线仅用于同一话题在不同时段的比较。
https://preview.redd.it/kiyjwvbgddyh1.png?width=3200&format=png&auto=webp&s=eef0df683049672427f6dbacdac29a17fbba81a5
https://preview.redd.it/d3sk40agddyh1.jpeg?width=3200&format=pjpg&auto=webp&s=1ded2efc1d3822ce96986ae738361580996d152d
https://preview.redd.it/55i5t1cgddyh1.png?width=3200&format=png&auto=webp&s=128cf93744dc06eb94f470bdced2c75519352ae0
I scored 342 US occupations on how exposed they are to AI and robots. Here's what the data shows and where the method is weak.
What it shows
- The top of the list is office and knowledge work. Customer service reps, general office clerks and software developers all score 94%. 60 jobs land in "Very High Risk", and together they employ 25.2 million people.
- The bottom is hands-on care. Home health aides, childcare workers and nursing assistants score 17%. Electricians (35%) and registered nurses (39%) also sit well below most office jobs.
- The riskier jobs pay more. The 60 Very High Risk jobs have a median salary of $79,665, against $51,030 for the 27 lowest-risk ones. That's the opposite of older automation waves, which hit lower-paid routine work first.
How it's built I started from Frey & Osborne's 2013 automation probabilities, adjusted each occupation for current AI capability with AI-assisted scoring, and joined BLS pay and headcount data.
Weak spots
- It scores whole occupations, not tasks. Most jobs will change long before they disappear.
- Scores only take 11 distinct values, so read them as bands, not precision.
- The "expiration year" is one rough year per job. A 2026 year means the change is already underway, not a layoff date.
- It's US-only.
Which job do you think it gets most wrong, and would a task-level approach fix that?
Sources: pay and headcount from the US BLS Occupational Outlook Handbook, and risk from Frey & Osborne (2013) plus AI-assisted adjustment, data from April 2026. The site and the free CSV are at aifiredyou.com . It's my own free, non-commercial side project.