AI assisting with specific work tasks inside a job, rather than replacing the role itself. Image credit: KorishTech (AI-generated).
Jobs most exposed to AI are concentrated in specific roles, according to the Anthropic AI labor report. The report identifies specific jobs where AI is already being used in real work, rather than estimating what it could do in theory. As explored in our previous analysis of the Anthropic AI labor report, this distinction reveals a gap between capability and actual usage.
The report shows that AI exposure is not evenly distributed. Instead, it is concentrated in occupations where a large share of daily tasks are already compatible with AI systems.
Which Jobs Are Most Exposed to AI
Anthropic finds that the highest exposure appears in digital, knowledge-based roles where work is performed through text, code, or structured data.
At the occupation level:
- Computer and mathematical roles → 35.8% observed exposure
- Office and administrative roles → 34.3%
- Business and financial roles → 28.4%
- Sales roles → 26.9%
Within these categories, certain jobs stand out with significantly higher task-level exposure:
- Computer programmers → ~70–75% of tasks
- Customer service representatives → ~70%
- Data entry roles → ~67%
- Medical record specialists → ~66%
- Market research analysts → ~65%
These roles share a common pattern. Their work is already structured in a way that AI systems can operate within.
What Makes a Job Highly Exposed
The difference between exposed and less-exposed jobs is not the job title, but the structure of tasks.
Jobs are highly exposed when their tasks are:
- text-based (emails, reports, code)
- structured and repeatable
- performed in digital environments
- already integrated into workflows where AI tools can be applied
For example, writing a report, generating code, or responding to customer queries can be broken into discrete steps that AI can assist or automate. These tasks resemble typical AI interactions, such as prompting, generating, and editing outputs.
In contrast, tasks that rely on physical action, real-time decision-making, or human interaction are less compatible with current AI systems.
Jobs With Low AI Exposure
Anthropic identifies a large share of occupations with little to no observed AI usage.
These include roles such as:
- cooks and food preparation workers
- mechanics and repair technicians
- construction and maintenance workers
- hospitality and service roles
These jobs remain less exposed because their tasks require:
- physical manipulation
- real-world context and environment
- direct human interaction
Even when some documentation or planning tasks exist, they represent a small portion of the overall role. As a result, the total exposure remains low.
Jobs Most Exposed vs Least Exposed
| Job Category | Theoretical Exposure | Observed Exposure | Task Type |
|---|---|---|---|
| Computer & Math (Programmers) | Very high (~90%+) | High (~70%+) | Coding, documentation |
| Office & Admin | High (~90%) | ~34.3% | Data entry, messaging |
| Business & Finance | High (~80%+) | ~28.4% | Reporting, analysis |
| Sales & Customer Support | High | ~26.9% | Communication, responses |
| Healthcare (Records) | High (partial) | ~66% (specific roles) | Documentation, coding |
| Physical & Service Jobs | Low (<20%) | Minimal | Manual, in-person work |
This pattern shows that exposure depends on how tasks are structured, not how important or skilled a job is.
Why High Exposure Does Not Mean Job Loss
High exposure indicates that many tasks within a job can be assisted or automated, but it does not mean the job itself disappears.
Anthropic’s data shows that:
- AI usage covers only a portion of tasks
- most use cases are still assistive, not fully automated
- unemployment has not increased significantly in exposed roles
This reflects the difference between capability and implementation. Even when AI can perform a task, it must be integrated into workflows before it affects employment.
What This Means in Practice
The early impact of AI appears in how work is organised rather than in immediate job loss.
Three patterns are emerging:
- tasks are being automated within roles
- hiring into exposed roles is slowing in some cases
- workflows are gradually incorporating AI tools
These changes are incremental. Instead of replacing jobs outright, AI is reshaping how those jobs function.
My Take
The list of “jobs most exposed to AI” can be misleading if interpreted as a prediction of job loss.
What the Anthropic report actually shows is a pattern of task-level compatibility. Jobs are exposed not because they are replaceable as a whole, but because parts of their work already fit into AI-driven workflows.
This explains why high exposure appears in roles like programming, customer service, and analysis. These jobs are built around tasks that can be structured, repeated, and processed digitally.
At the same time, this does not translate directly into job replacement. As long as only part of the workflow is automated, the role continues to exist, but with a different internal structure.
The more useful way to interpret exposure is as a signal of where change is happening first. Jobs with high exposure are likely to evolve faster, with shifts in task composition, expectations, and required skills.
This suggests that the impact of AI is not evenly distributed. It begins in specific types of work and expands outward as workflows adapt, rather than affecting all jobs at once.
Sources
Anthropic — Labor market impacts of AI: A new measure and early evidence
https://www.anthropic.com/research/labor-market-impacts
Anthropic — Anthropic Economic Index
https://www.anthropic.com/economic-index
Eloundou et al. (2023) — GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models
https://arxiv.org/abs/2303.10130