The Slow Revolution: AI Skills Demand in U.S. Healthcare Job Postings Exploring Trends in Healthcare (Hospitals Subsector) AI Adoption

Artificial intelligence reduces the cost of prediction, yet its adoption inside hospitals has proceeded far more slowly than in finance or professional services. This paper investigates why by measuring explicit AI skill requirements in 13,843,830 U.S. hospital job postings from 2,432 employers spanning 2015 to 2023.Using the Burning Glass Technologies (BGT) vacancy dataset and a conservative, time-stable keyword definition, the analysis shows that only about one in a thousand hospital postings (0.11%) explicitly requires AI skills.The central finding is that adoption is organizationally situated: IT and research roles are roughly nine to twelve times more likely to list AI skills than administrative roles (the reference category), while clinical and teaching positions remain well below the baseline.A monotonic educational gradient further reveals that PhD-level postings are sixteen times more likely to demand AI than high-school-level ones.At the employer level, a 10 percentage-point higher research posting share is associated with a 0.13 percentage point higher AI posting share — a sizable effect relative to the 0.1 to 0.2 percentage point typical hospital mean.These patterns hold across alternative keyword definitions, sample restrictions, and clustering assumptions.Together, the findings point to organizational and regulatory frictions — not the absence of useful AI applications — as the principal barriers to broader, clinically situated diffusion, and suggest that complementary investments in analytics workforce development and governance capacity are necessary preconditions for meaningful hospital AI adoption at scale.