Across 342 occupations, measured AI exposure varies more between domains than within them. Marketing & Sales occupations carry the highest mean exposure (85th percentile across 10 roles), and Trades & Healthcare the lowest (29th across 140). Exposure is measured from Microsoft Research and Anthropic Economic Index telemetry. Domain sizes are uneven — they range from 10 occupations to 140 — so a mean drawn from a small group is a weaker signal than one drawn from a large one.
Exposure alone is the less interesting half. What separates these domains is how differently that exposure resolves: of the 342 occupations, only 47 are both exposed and classed by Goldman Sachs as substituted, against 202 where AI reaches just as deep but amplifies the work instead. In Technology, all 43 occupations fall on the amplifying side, none in displacement risk. In Business, Finance & Legal it inverts: 27 of 29 sit in displacement risk.
Ordered by mean measured exposure, highest first. Percentile is the mean of the measured exposure composite across occupations in that domain that carry one; the automation and reshaping columns are modelled estimates.
| Domain | Roles | Mean exposure | Displacement | Orchestrator | Insulated | Modelled auto. / reshape |
|---|---|---|---|---|---|---|
| Marketing & Sales | 10 | 85th | 8 | 2 | 0 | 56% / 75% |
| Business, Finance & Legal | 29 | 69th | 27 | 1 | 1 | 43% / 64% |
| Technology | 43 | 67th42/43 measured | 0 | 43 | 0 | 41% / 63% |
| Science & Education | 43 | 67th42/43 measured | 0 | 42 | 1 | 42% / 63% |
| Creative & Media | 27 | 65th | 0 | 25 | 2 | 40% / 61% |
| Management & Ops | 50 | 54th47/50 measured | 12 | 26 | 12 | 33% / 53% |
| Trades & Healthcare | 140 | 29th139/140 measured | 0 | 63 | 77 | 17% / 33% |
10 occupations, mean measured exposure 85th percentile. 8 sit in displacement risk, 2 in orchestrator opportunity, and 0 are insulated. On modelled estimates, an average of 56% of tasks are already automated and 75% reshaped rather than replaced.
Most exposed in this domain29 occupations, mean measured exposure 69th percentile. 27 sit in displacement risk, 1 in orchestrator opportunity, and 1 are insulated. On modelled estimates, an average of 43% of tasks are already automated and 64% reshaped rather than replaced.
Most exposed in this domain43 occupations, mean measured exposure 67th percentile. 0 sit in displacement risk, 43 in orchestrator opportunity, and 0 are insulated. On modelled estimates, an average of 41% of tasks are already automated and 63% reshaped rather than replaced.
Most exposed in this domain43 occupations, mean measured exposure 67th percentile. 0 sit in displacement risk, 42 in orchestrator opportunity, and 1 are insulated. On modelled estimates, an average of 42% of tasks are already automated and 63% reshaped rather than replaced.
Most exposed in this domain27 occupations, mean measured exposure 65th percentile. 0 sit in displacement risk, 25 in orchestrator opportunity, and 2 are insulated. On modelled estimates, an average of 40% of tasks are already automated and 61% reshaped rather than replaced.
Most exposed in this domain50 occupations, mean measured exposure 54th percentile. 12 sit in displacement risk, 26 in orchestrator opportunity, and 12 are insulated. On modelled estimates, an average of 33% of tasks are already automated and 53% reshaped rather than replaced.
Most exposed in this domain140 occupations, mean measured exposure 29th percentile. 0 sit in displacement risk, 63 in orchestrator opportunity, and 77 are insulated. On modelled estimates, an average of 17% of tasks are already automated and 33% reshaped rather than replaced.
Most exposed in this domainA domain average describes a group of occupations, not any one job and certainly not any one person. The spread inside a domain is often wider than the gap between domains — which is the argument for looking up your own occupation rather than your industry. How this is measured · Find your occupation
An occupation average tells you where the ground is moving, not where you stand on it. Map the skills you actually own against it, and see which pathways stay open.