In April 2026 the Bipartisan Policy Center launched an AI and Workforce Navigator with a live Skills Data Dashboard powered by Lightcast, which scrapes millions of US job postings and tags any that ask for one of more than 300 AI-related skills. The first analysis, in May, found AI-skill postings up 144% year on year. The September update, published this week, puts the figure at 165%.
Over the same period, overall job postings grew about 7%, essentially flat.
Key takeaways:
- +165%: year-on-year growth in US job postings requiring AI skills (September 2026 update). It was +144% in April.
- +7%: growth in overall job postings over the same period.
- Demand is no longer concentrated in tech; professional-services industries now show similar or faster growth.
- In accountancy offices, the fastest-growing AI skills are Microsoft Copilot (+85%), prompt engineering (+76%), MLOps (+67%), retrieval-augmented generation (+64%) and generative AI (+60%).
- Gusto data cited in the September update: unique AI-related job titles at small businesses rose from 70 in 2019 to nearly 600 today.
How Lightcast counts
Lightcast maintains a taxonomy of around 33,000 skills. From it, the firm has identified 300+ AI-related skills organised into clusters such as generative AI, machine learning, natural-language processing, robotics, and AI ethics, governance and regulation. A posting counts as "requiring AI skills" if it mentions any one of them, whether that is "gradient boosting" or "prompt engineering".
The method is used by researchers at Stanford, the OECD and PwC, so figures from those sources are broadly comparable.
The shift beyond tech
The July analysis is the most useful for career planning, because it breaks demand down by industry. The finding: the skills accountants, bankers and staffing firms now advertise for are the same advanced AI skills previously associated with technical roles.
| Skill (accountancy offices) | Year-on-year growth in postings |
|---|---|
| Microsoft Copilot | +85% |
| Prompt engineering | +76% |
| MLOps | +67% |
| Retrieval-augmented generation (RAG) | +64% |
| Generative AI | +60% |
The pattern tells a coherent story. Copilot and prompt engineering reflect demand for people who can direct AI tools. MLOps and RAG reflect demand for people who can maintain and ground them once deployed. Both are appearing in job ads for firms that have never employed a data scientist.
A caution the BPC itself raises
Job postings show what employers ask for, not what they use. The September update notes that some sectors still lag in actual AI adoption even as their postings fill with AI terms. Demand signals and usage can diverge, particularly where firms are hiring ahead of capability.
Concept to take away: skills, not titles. The dashboard tracks skills because titles are unreliable: Gusto counts nearly 600 distinct AI-related job titles at small businesses alone. When you plan a learning path, anchor on named, checkable skills (RAG, MLOps, prompt engineering, model evaluation) that appear in postings across industries, not on a title that may not exist next year.
Why learners should care
Two things. First, the growth rate is the most current public measure of AI-skill demand and it is still accelerating. Second, it confirms that AI skills are becoming a cross-industry requirement, which means a career changer in finance, healthcare or manufacturing does not need to become a software engineer to benefit.
