Most Companies aren't Using AI Agents Yet. That's the Story.
The gap between the AI agent narrative and the actual adoption data is itself the most important thing to understand about this market right now.
McKinsey's 2025 State of AI Global Survey — the most granular current reading on enterprise agentic AI — found that 23% of organizations are actively scaling an AI agentic system in at least one business function, and another 39% have begun experimenting. That sounds like meaningful penetration until you look at the next level of detail: in any given business function, no more than 10% of respondents report their organizations are scaling AI agents. Most of those that are scaling are doing so in only one or two functions.
Aaron Levie, CEO of Box, posted on X this month that "the majority of companies aren't even using coding agents at scale, let alone for the rest of knowledge work." The McKinsey data backs him up. The 23% figure means that roughly one in four organizations has put agents into production somewhere in the business — not most departments, not most workflows, somewhere.
The sector breakdown from the survey tells a more specific story. Technology companies lead in software engineering and IT. Insurance leads in marketing and sales. Healthcare shows strength in knowledge management and IT. These aren't horizontal adoption curves — they're vertical deployments solving specific workflow problems in specific industries.
The capital tells a similar story. CB Insights data shows AI startups absorbed more than half of all global VC dollars in the first half of 2025, with H1 total hitting $116 billion — already exceeding all of 2024. But capital deployed is not the same as value created. The McKinsey data shows that while 64% of respondents say AI is enabling innovation at their companies, only 39% report EBIT impact at the enterprise level. The money moved faster than the ROI.
The cloud analogy is instructive. In 2010, the entire technology industry had converged on the conviction that cloud was inevitable. AWS generated just $500 million in revenue that year. By 2025, AWS, Azure, and Google Cloud combined for revenues approaching $400 billion annually — roughly an 800-fold expansion from a base the industry already believed was the future. The companies that bought early waited. The companies that built infrastructure to serve the buyers made the money.
The McKinsey data suggests we're somewhere around 2010 in the agentic AI adoption curve. 23% have something in production. 10% at the function level. The gap between the narrative and the data is real — and for founders and builders, it's the most important signal in the room. The enterprise is intrigued, cautious, and still figuring out where agents actually reduce cost versus where they introduce new categories of risk.
McKinsey's State of AI Trust in 2026 report adds one more dimension: only about 30% of organizations have reached a maturity level of three or higher in strategy, governance, and agentic AI controls. Asia-Pacific leads globally. Technology, media, telecommunications, and financial services outperform other sectors. The countries and companies moving fastest on agents are also the ones building the compliance and governance infrastructure first — which suggests the organizations likely to capture value from agents at scale are the ones treating it as an infrastructure problem, not a chatbot problem.
For founders and builders: the enterprise buyer is real, they're engaged, and most of them haven't figured out where agents fit yet. That's the opportunity. The companies that help enterprises answer the governance and controls question — not just the efficiency question — are the ones positioning for the next phase.
Primary sources:
McKinsey 2025 State of AI Global Survey, November 5, 2025
Forbes — "Roughly 10% Of Enterprise Functions Use AI Agents, McKinsey Finds", March 22, 2026
McKinsey — State of AI Trust in 2026: Shifting to the agentic era
