Forty-seven per cent of architecture and engineering work needs software built first, and 19 per cent is reachable today. The protected third is the part that happens on a site.
Exposure means published research judged an AI could halve the time a task takes, at the same quality. It is not a prediction about any job. How to read this
The work is analysis, modelling, process design and documentation, which is close to pure text and numbers. Automotive Engineering Technicians are the most protected at 67 per cent, because it involves testing physical things.
Fifty-seven points of spread between two jobs with "engineering" in the title.
Engineering already has decades of purpose-built software behind it: modelling, simulation, drafting, project control. Yet 47 per cent of the work still sits in the band rated as needing a tool.
Which suggests the band is not a measure of whether software exists in a field. It is a measure of whether software exists for that specific task.
Solid is reachable with a chat window today, pale needs software building first, grey cannot be sped up. Task counts are shown because several of these rest on fewer than twenty.
Engineering has had specialist software for forty years and still shows a 47 per cent middle band. Accounting is the same. The lesson for People teams is that "there must be a product for this already" is usually wrong at the task level.
The gap is rarely a whole category of software. It is one specific job nobody has bothered to automate because it belongs to somebody who cannot build.
There is no data here on what anybody actually does with AI. A dataset that appears to measure it was removed from this report because its meaning could not be stated. The reasoning is on the limits page.
The exposure ratings cover all 1,257 tasks in these 56 occupations with no gaps, and were made in 2023 against the model capabilities of that year. Read every figure as a floor.
Exposure ratings from Eloundou, Manning, Mishkin and Rock, GPTs are GPTs: Labor market impact potential of LLMs, Science 384, 1306–1308, 2024, used under the MIT licence. This page includes information from the O*NET Database by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA), used under the CC BY 4.0 license. O*NET® is a trademark of USDOL/ETA. People Team AI has modified all or some of this information. USDOL/ETA has not approved, endorsed, or tested these modifications. Employment figures from the US Bureau of Labor Statistics.