Enterprise AI Is Everywhere. Agents? Still Rarer Than the Hype Suggests.

Here's the number that cuts through the agent hype: fewer than 10% of organizations have scaled AI agents in any individual business function. That's according to McKinsey's 2025 State of AI survey, the annual report based on responses from 1,993 participants across 105 countries — the closest thing the industry has to a ground-truth barometer. The finding was reported by Forbes this week, and it deserves the attention because it's more honest than most of what passes for agent news.

The paradox at the center of the data is the real story. AI adoption is genuinely broad: 88% of organizations now deploy AI in at least one function, up from 78% the prior year. More than two-thirds use AI across multiple functions simultaneously — IT operations, marketing, customer service, knowledge management, product development. By that measure, the enterprise AI era has arrived. But only about one-third have genuinely scaled AI across the enterprise. The rest are in what analysts call the pilot loop: proof-of-concept launches, localized wins, and no path to systemic deployment.

The agent numbers are even more striking. Most organizations that are scaling agents are doing so in only one or two functions. In any given function — take HR, or finance, or supply chain — fewer than 10% of respondents say their organization is actively scaling AI agents. This is not a technology problem. The tools exist. The constraints are organizational: legacy data infrastructure, workflows that weren't redesigned for AI execution, and a lack of clear scaling priorities that spreads investment across dozens of initiatives instead of concentrating it.

Where agents are actually working, the ROI numbers are real. Software engineering and IT functions report 10–20% cost reductions from agent deployment. Marketing and product development teams are seeing revenue uplifts above 10%. Per McKinsey's separate agent-economics analysis — "Agents for growth: Turning AI promise into impact," published November 3, 2025 — effective and scaled agent deployments could deliver productivity improvements of 3–5 percent annually and potentially lift growth by 10 percent or more. Those figures are forward-looking and conditional on actually reaching scaled deployment, which is precisely what most organizations haven't done yet. The gap between those numbers and the 10% scaling figure is the opportunity — and the warning.

Only 6% of companies qualify as what McKinsey calls "high performers": organizations where AI contributes meaningfully — more than 5% — to EBIT in a lasting way. The distinguishing factor is not budget or headcount. It's senior leadership ownership, long-term commitment, and an operating model that rewires processes for AI rather than simply layering AI onto existing workflows. High performers treat AI as infrastructure, not as a project.

The contrast with what's happening in China is instructive. While enterprise adoption in the West sits at under 10% per function, Chinese tech companies are racing to embed agent capabilities into their core platforms at speed — Tencent wiring OpenClaw into WeChat, Alibaba launching a multi-agent enterprise platform, Baidu putting agents across its hardware ecosystem. The adoption curve there is being driven by platform distribution and consumer scale, not by careful enterprise rollout. Different risks, different timeline, different governance model.

Our read: the McKinsey number is not a knock on agents — it's a realistic baseline for where enterprise infrastructure stands today. The 10% is actually encouraging if you believe in the ROI data, because it means 90% of the opportunity is still in front of the organizations that figure out how to exit the pilot loop. The tools work. The economics work in specific functions. The blocker is organizational design, not technology readiness. The companies that solve for that — that treat agent infrastructure as a rewiring problem rather than a deployment problem — are the ones who will close the gap over the next 18 to 24 months.