Organizations that deploy generative AI with a defined scope and a measurable baseline see a median 3.7 times return on their investment, according to McKinsey's State of AI research. The finding stands out against a broader pattern in which fewer than one in ten enterprises can point to a generative AI deployment delivering measurable, sustained value at scale.

Adoption itself is widespread. McKinsey reports that 78 percent of organizations now use AI in at least one business function, up from 55 percent just two years earlier. The firm estimates $2.6 trillion to $4.4 trillion in annual addressable value globally, a ceiling that most companies have yet to approach.

Function-level data show where returns concentrate. In customer operations, organizations deploying AI copilots achieved a 30 to 45 percent reduction in average handle time, with first-contact resolution rates rising 15 to 25 percentage points when AI agents grounded in company data handle basic volume. In software engineering, developers using AI code assistants completed tasks 25 to 40 percent faster, and code review cycles shrank by roughly 30 percent.

McKinsey attributes most failures to organizational factors rather than technology: projects scoped too broadly, data architectures never built for production, change management treated as an afterthought, and a lack of baseline metrics needed to prove returns.

Source: McKinsey -- https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai