Which Jobs Are Most Vulnerable to AI? A Data-Driven Look at Exposure and Growth

By shrijeetverma13 · August 3, 2026

We analyzed how artificial intelligence exposure relates to job growth, wages, and required skills across hundreds of occupations. The findings reveal…

The analysis compared projected employment growth (2024-2034) across Low, Moderate, and High AI exposure categories, generating data tables that break down average and median growth rates for each group. While the specific numeric breakdown isn't fully displayed here, the underlying data tables capture the count of occupations and their growth statistics within each AI exposure level, allowing for direct comparison of how AI exposure relates to job growth projections.

The analysis grouped occupations by required education level to compare average median wages and AI exposure scores. Two data tables were generated showing these breakdowns across different education categories, from no formal credential through doctoral degrees, along with the number of occupations in each group.

Among top-quartile earners (median wage ≥ $93,955), several occupations stand out for combining high pay with very low AI exposure. Optometrists lead with $134,830 wage and just 0.050 AI exposure, followed by Elevator/Escalator Repairers, Physician Assistants, Dentists, Power Plant Operators, Radiation Therapists, Airline Pilots, and Veterinarians. These careers span diverse education levels, from high school diplomas to doctoral degrees, showing that resilience to AI isn't limited to highly credentialed fields.

The analysis identified which job categories face the greatest AI exposure and employment risk by combining AI exposure scores (from language model-based metrics) with projected employment growth data. Two visualizations were generated: one ranking job categories by their average AI exposure level (color-coded by growth rate), and another ranking categories by a combined employment risk score that factors in both high AI exposure and weak or negative growth outlook. Categories with high exposure and low growth—such as many office and administrative-type roles—tend to show up as the most at-risk, while categories with high exposure but strong growth are exposed to AI but less likely to see job losses.

13 occupations combine negative projected growth (2024-2034) with high AI exposure. The most vulnerable is Desktop publishers, projected to shrink by 12% with a notable AI exposure score. Other high-risk roles include General office clerks (-7% growth), Advertising sales agents (-6%), Survey researchers (-5%), and Customer service representatives (-5%). Many of these are office, administrative, and media-related jobs where AI tools can increasingly automate core tasks.

Occupations where Spatial abilities are the distinctive cognitive strength have the lowest average AI exposure (0.199 on the LLM scale), across 66 occupations. In contrast, Quantitative reasoning strengths show the highest AI exposure (0.504) across 48 occupations. Overall, the ranking from lowest to highest AI exposure is: Spatial, Perceptual, Attention, Memory, Reasoning, Verbal, and Quantitative. This suggests that jobs relying on spatial, perceptual, and attention-based skills are currently less exposed to AI/LLM automation, while jobs built around verbal and quantitative reasoning are more exposed, since large language models are especially strong at text and numeric tasks.

The analysis grouped occupations by required education level and produced summary tables showing average AI exposure, projected growth percentage, and occupation counts for each education tier. While the specific correlation values and charts encountered a technical hiccup during rendering, the underlying grouped data tables were successfully generated and are available for review, capturing how AI exposure and job growth outlook vary across education requirements from no formal credential up through doctoral degrees.

The analysis compared three AI exposure metrics with seven cognitive ability scores. Verbal ability showed the strongest positive relationship with AI exposure, especially with the AIOE metric (r=0.808). In contrast, spatial ability showed the strongest negative relationship, particularly with AIOE (r=-0.454). This suggests jobs or tasks requiring more verbal skills tend to have higher AI exposure, while those relying more on spatial skills tend to have lower AI exposure. A heatmap was created to visualize these correlations across all combinations of AI exposure metrics and cognitive abilities.

Using the IQR method, only one occupation stood out as an outlier in both median annual wage and employment levels: Financial Managers. This role has a median annual wage of $161,700 and employs 868,600 people, making it exceptionally high in both pay and workforce size compared to other occupations in the dataset.

The analysis normalized both the AIOE (task-based) and LLM-Human AI exposure scores to a 0-1 scale and identified the 10 occupations where these two measures disagree the most. Two visualizations were generated: a grouped bar chart comparing the normalized AIOE and LLM-Human scores for each of the top 10 divergent occupations, and a second chart summarizing which job categories these occupations fall into. Several data tables with the detailed rankings and category breakdowns are also available for review. These divergent occupations tend to cluster in a few dominant job categories, suggesting that AIOE (which is based on task structure) and the LLM-Human rating (which reflects language-model relevance) capture fundamentally different dimensions of AI exposure — some jobs look highly automatable by one measure but not the other.