AI-first company

Assessment · Task and job maps

AI Occupational Exposure measure (AIOE)

Which of our roles could AI touch most?

By
Last reviewed
Corrections
Report an error
Who answers Nobody is surveyed
How long You match your job titles to United States occupation codes
Can you use it? Free to use, ask before changing
Testing Moderate

Tested properly, but by one team or in one setting. Nobody independent has yet repeated the test.

In short

A public score of how far each occupation relies on abilities where AI has advanced. Matching scores exist for industries and regions. Use it to compare roles at a glance, then look at real tasks.

For whoever runs the assessment: an HR or operations lead, an internal project team, or an outside consultant.

What it measures

Each occupation needs a mix of abilities. AI has advanced in some of them and not in others. The measure links the two and gives each occupation one exposure score. The authors also built scores for industries and for regions, and show how to build one for a single firm from the mix of occupations it employs.

The study is described on its own page.

How to use it

  1. Match your roles to the occupation codes.
  2. Use the scores to rank roles for a closer look.
  3. Map the tasks in the top few roles with the people who do them. See map tasks before deciding what to automate.
  4. Do not present the scores to staff as a forecast.

What it does not measure

  • Whether AI will replace, assist or leave the work alone. It gives one number for exposure and does not separate those cases.
  • Tasks. It scores whole occupations. Two people with the same title can do different work.
  • Use. It shows what AI could touch, not what it is used for.
  • Your country’s job list. It uses United States codes.

How it was tested

  • The authors built the measure and checked it in a peer-reviewed study.
  • A separate team applied it to 23 countries. More exposed occupations did not lose employment overall. Where computer use was high, more exposure went with faster job growth.
  • The authors later extended it to AI language tools in a working paper. That paper has not been peer-reviewed.

So the score is well documented, and the best-known test of it is a warning against reading it as a forecast of job losses.

Licence and access

The article is open access under CC BY 4.0. The authors share the data. Its location and licence are not established.

Alternatives

Other assessments with a similar job, and when to prefer each.

Practices that use it

The details

Testing
Moderate Built and checked by its authors in a peer-reviewed study. A separate team applied it across 23 countries and found that exposure did not predict job losses. It has not been tested as a guide to decisions inside one company.
Looks at
Job and its tasks
How the answers are collected
Existing data on the abilities each occupation needs, linked to areas where AI has advanced.
Length
No survey. You match your job titles to United States occupation codes.
Cost
Free
Licence
CC BY 4.0 (article)The article is openly licensed. The terms for the data set are not established.
Languages
English. Built on United States job classifications; other countries need a mapping.
By
Felten, Raj and Seamans, 2021

Sources

  1. Felten, Raj and Seamans (2021). Occupational, industry, and geographic exposure to artificial intelligence: A novel dataset and its potential uses. Strategic Management Journal. doi:10.1002/smj.3286
  2. Felten, Raj and Seamans (2023). How will Language Modelers like ChatGPT Affect Occupations and Industries? Working paper, not peer-reviewed. doi:10.2139/ssrn.4375268
  3. Georgieff and Hyee (2022). Artificial Intelligence and Employment: New Cross-Country Evidence. Frontiers in Artificial Intelligence. doi:10.3389/frai.2022.832736

Related studies on this site