33.4% of US Sales Jobs Are in Occupations With High AI Exposure

Hollerly analysis of BLS OEWS May 2025 employment and published GPT exposure scores

Key findings

• 33.4% of US sales jobs (4,666,000 of 13,968,180) are in occupations where at least half of tasks are exposed to large language models, based on GPT-4 ratings.
• Telemarketers are the most AI-exposed sales occupation, with an exposure score of 0.81.
• Demonstrators and Product Promoters are the least exposed sales occupation, at 0.13.
• 83.1% of workers in B2B-focused sales roles are in high-exposure occupations, compared with 33.4% of all sales workers.
• Sales workers in high-exposure occupations earn an average of $84,680, compared with $47,513 in lower-exposure sales occupations.
• Using human annotators' ratings instead of GPT-4's, 31.6% of sales jobs fall in high-exposure occupations.

What "exposure" means here

Exposure is not job loss. The scores used in this study come from Eloundou, Manning, Mishkin and Rock, "GPTs are GPTs: Labor market impact potential of LLMs," published in Science in 2024. The authors rated every task in the O*NET database on whether access to a large language model, alone or with additional software built on top of it, could cut the time needed to complete the task by at least half. The occupation-level beta score used here counts tasks directly exposed to the model plus half of the tasks that need additional software. A score of 0.5 means roughly half of an occupation's tasks are exposed.

An exposed task can be sped up, reshaped or handed off. Whether that leads to more output per rep, different job content or fewer roles depends on employers and markets, not on the score.

Which sales jobs are most exposed

Telemarketers top the list at 0.81. The B2B-focused roles cluster in the middle and upper part of the range: the employment-weighted exposure for those seven occupations is 0.58, compared with 0.30 for retail sales occupations.

How many workers this covers

Of the 13,968,180 sales workers in the 22 occupations with exposure scores, 4,666,000 (33.4%) work in occupations scored 0.5 or higher. The largest sales occupations by headcount, retail salespersons and cashiers, score below the threshold, which keeps the overall share down.

Exposure and pay

The more exposed sales occupations tend to be the higher-paid roles. Workers in high-exposure sales occupations average $84,680 a year, against $47,513 for the rest.

Full table

Hollerly's AI SDR software researches leads, writes LinkedIn and email outreach and drafts replies, with a person approving every campaign.

About the data

• Exposure scores: Eloundou, T., Manning, S., Mishkin, P., and Rock, D. (2024). "GPTs are GPTs: Labor market impact potential of LLMs." Science 384(6702): 1306-1308. Preprint: arXiv:2303.10130. Occupation-level scores as published in the authors' public data repository.
• Measure: the beta score (E1 + 0.5 x E2) from GPT-4 ratings is the primary measure. Human annotator ratings of the same tasks are reported as a check; the two measures have a correlation of 0.64 across these occupations. "High exposure" is our threshold of 0.5 or higher, meaning at least half of an occupation's tasks are rated exposed.
• Employment and wages: U.S. Bureau of Labor Statistics, Occupational Employment and Wage Statistics (OEWS), May 2025 national estimates, the latest release available as of 2026-10-02.
• Sample: all detailed occupations in SOC major group 41 (sales and related) plus 11-2022 sales managers, matched on 6-digit SOC to the O*NET ".00" occupation code. 1 occupation without an exposure score is not included: Sales and Related Workers, All Other (93,180 jobs). The "B2B-focused" subset is SOC 11-2022, 41-1012, 41-3011, 41-3091, 41-4011, 41-4012 and 41-9031.
• Limitations: exposure measures potential task change, not employment effects. The scores were built on 2023-era model capabilities and O*NET task lists. OEWS excludes self-employed workers.

Sources

• Eloundou et al. (2024), Science: science.org/doi/10.1126/science.adj0998
• arXiv preprint: arxiv.org/abs/2303.10130
• Exposure data: github.com/openai/GPTs-are-GPTs/blob/main/data/occ_level.csv
• BLS OEWS May 2025 national estimates: bls.gov/oes/special-requests/oesm25nat.zip