McKinsey’s State of AI survey (November 2025) shows another jump in reported use—and the same hard gap between pilots and scaled value. Eighty-eight percent of respondents say their organizations regularly use AI in at least one business function, up from 78% a year earlier. About one-third have begun scaling. Enterprise EBIT impact remains uncommon.
Adoption is no longer the story
Regular use in at least one function is now the default among surveyed organizations. The interesting split is stage: most remain in experimenting or piloting at the enterprise level, while roughly one-third say they have begun to scale.
That framing matters for planning. Buying more seats is not the same as redesigning a workflow and measuring outcomes.
EBIT impact is still rare
Thirty-nine percent attribute any level of enterprise EBIT impact to AI, and most of those put it below 5%. McKinsey’s “high performers”—about 6%—report at least 5% EBIT impact and significant overall value. They are more likely to redesign workflows, scale faster, and invest more heavily.
Use-case cost and revenue benefits show up more often than company-wide EBIT. Innovation and customer satisfaction are frequently cited even when the P&L line is still thin.
Agents are early—and narrowly deployed
Interest in agentic systems is real: many organizations are experimenting, and a smaller share are scaling agents somewhere in the enterprise. But in any given business function, fewer than 10% of respondents say their organizations are scaling AI agents.
Demos that imply agents are already ubiquitous overstate the field. Tooling, permissions, and reliability still dominate the work.
What to take from the survey
Treat “we use AI” as table stakes. Ask whether a named workflow has owners, acceptance criteria, and a KPI that would show up in operations—not only in a slide deck.
If you are not yet in the high-performer group, the survey’s signal is not “try more models.” It is “scale fewer things that change how work gets done.”