CONCEPT
Deployment Overhang
The gap between what frontier AI systems are capable of doing and what users, organizations, and institutions have learned to do with them — a lag period during which most of the value (and most of the risk) of a new capability sits unrealized.
Deployment overhang is the condition in which AI capabilities exist in the wild but have not been integrated into workflows, products, or institutions well enough for their effects to be felt. The term is a relative of "compute overhang" (more compute exists than is being used to train frontier models) and "hardware overhang" (more capable hardware exists than algorithms to saturate it), but describes a different gap: between what is technically possible today and what is operationally realized. In 2025 the deployment overhang for large language models is vast. A large majority of knowledge workers have not meaningfully integrated frontier AI into their daily work despite the capability being free at a generous tier. The gap is closing, but slowly relative to the rate of capability advance.
In The You On AI Field Guide
The evidence for the overhang is empirical. Surveys of knowledge-work adoption (McKinsey 2024, Anthropic's own economic-impact