AI’s Australian Job Shake-Up: Anthropic’s Exposed Roles

The rapid advancement of Artificial Intelligence (AI) is undeniably reshaping the global employment landscape, sparking widespread discussion and concern. As organisations, academics, and businesses grapple with the potential ramifications, a new report from a leading AI developer offers a fresh perspective on which jobs are most susceptible to automation.

Anthropic, the creators of the AI model Claude, has released a comprehensive study titled “Labour Market Impacts of AI: A New Measure and Early Evidence.” This report distinguishes itself by not only examining the theoretical capabilities of AI but also by analysing real-world usage data to gauge “observed exposure.” This crucial distinction highlights the difference between what AI could do and what it is actually doing in the workplace.

Understanding Observed Exposure vs. Theoretical Capability

The core of Anthropic’s research lies in its introduction of the “observed exposure” metric. While theoretical capability quantifies the tasks an AI could theoretically accelerate or automate, observed exposure measures the actual extent to which these AI tools are being implemented and used in practice. This practical application is a far more accurate indicator of immediate job displacement risk.

The report visually represents this concept, with the “red area” illustrating how professionals are currently using Claude in their work. The researchers posit that as AI capabilities grow and adoption deepens, this red area is expected to expand, eventually encompassing the “blue area” which represents theoretical potential. However, they also acknowledge that significant portions of work, particularly those involving manual labour (like agricultural tasks or operating heavy machinery) and complex human interaction (such as courtroom representation), remain largely beyond AI’s current reach.

Sectors with High Theoretical AI Coverage

Based on Anthropic’s analysis of 22 occupational groups, several sectors demonstrate a substantial theoretical capacity for AI integration:

• Computer and Math Occupations: These fields show the highest theoretical AI coverage, standing at an impressive 94.3%.
• Business and Finance Occupations: Mirroring the tech sector, these roles also boast a theoretical AI coverage of 94.3%.
• Management: This broad category shows significant theoretical potential at 91.3%.
• Office and Administrative Support: With 90% theoretical coverage, these roles are highly amenable to AI assistance.
• Legal Occupations: Theoretical AI coverage in the legal field reaches 89%.
• Architecture and Engineering: These professions have a theoretical AI coverage of 84.8%.
• Arts and Media: This creative sector also shows substantial theoretical potential at 83.7%.

Beyond these, five additional occupational groups show over 50% theoretical scope for Large Language Model (LLM) penetration:

• Life and Social Sciences (77%)
• Sales (62%)
• Education and Library Occupations (61.7%)
• Healthcare Practitioners (59.9%)
• Social Services (50.5%)

Sectors with Lower Theoretical AI Coverage

Conversely, some sectors appear to have considerably less theoretical room for AI integration. These include:

• Ground Maintenance: This sector has the lowest theoretical AI coverage at just 3.9%.
• Transportation: Only 12.1% of jobs in this sector are theoretically open to AI usage.
• Agriculture: Theoretical AI coverage stands at 15.7%.
• Food and Serving: This sector has 16.9% theoretical AI coverage.
• Construction: Similar to food and serving, construction shows 16.9% theoretical coverage.
• Personal Care: This area has 18.2% theoretical AI coverage.
• Installation and Repair: Theoretical AI coverage is 18.4%.
• Production: This sector shows 19% theoretical AI coverage.

Other sectors with lower theoretical AI coverage include Healthcare Support (28.5%) and Protective Services (31.6%).

Observed Exposure: Where AI is Making Real Inroads

While theoretical potential is one thing, actual implementation is another. The Anthropic report highlights the following sectors with the highest observed AI exposure, indicating a more immediate risk of displacement:

• Computer and Math Occupations: These lead with the highest observed AI coverage at 35.8%.
• Office and Administrative Roles: Closely following, these roles show 34.3% observed AI coverage.
• Business and Finance: This sector has an observed AI coverage of 28.4%.
• Sales: Sales roles demonstrate 26.9% observed AI coverage.
• Legal Occupations: Observed AI exposure is 20.4%.
• Arts and Media: This sector has 19.2% observed AI coverage.
• Education and Library Occupations: These roles show 18.2% observed AI coverage.

The Ratio: How Much Potential is Being Realised?

The ratio of observed exposure to theoretical capability provides a crucial insight into how much of the potential AI automation is already being put into practice.

• Sales: This sector leads with a ratio of 43% (meaning 27% of sales jobs are exposed, out of a theoretical potential of 62%).
• Office and Administrative Jobs: These roles have a ratio of 38%.
• Computer and Math Occupations: Also showing a 38% ratio.
• Business and Finance: This sector’s ratio is 30%.
• Education and Library Occupations: These fields also have a 30% ratio.

Interestingly, while architecture and engineering have a very high theoretical capability (85%), their observed exposure ratio is only 5%, suggesting a slower pace of actual AI adoption in these fields thus far.

Individual Occupations Most Exposed

Zooming in on specific job roles, the report identifies several occupations with the highest observed AI exposure:

• Computer Programmers: This occupation shows the highest observed AI exposure at a significant 74.5%.
• Customer Service Representatives: At 70.1%, these roles are highly exposed.
• Data Entry Keyers: Showing 67.1% observed AI exposure.
• Medical Record Specialists: This role has 66.7% observed AI exposure.
• Market Research Analysts and Marketing Specialists: These professionals face 64.8% observed AI exposure.
• Sales Representatives (Wholesale and Manufacturing, excluding technical/scientific products): This group has 62.8% observed AI exposure.

Demographics of Exposure and the Unemployment Question

The data also offers insights into the demographics of workers in the most exposed professions. Generally, those in the highest-risk occupations tend to be older, possess higher levels of education, earn more, and are more likely to be women.

Crucially, despite this observed exposure, the report found no systematic increase in joblessness among workers in heavily exposed occupations since late 2022. However, there is suggestive evidence indicating a slowdown in the hiring of younger workers within these same fields. This is a trend that warrants close monitoring as AI integration continues to evolve.