Nearly every company now uses AI, but most still can't turn pilots into profit

Three years after generative AI tools arrived, AI use in business has become near-universal. In the 2025 McKinsey Global Survey, 88% of respondents say their organizations regularly use AI in at least one business function, up from 78% a year earlier and just 20% in 2017. More than two-thirds report AI use across multiple functions, and half now use it in three or more.

Yet the gap between adoption and impact remains stark. Among organizations using AI, 32% are still experimenting, 30% are piloting, and only 31% have begun scaling. A mere 7% describe AI as fully deployed and integrated across the enterprise. Larger companies lead the way: nearly half of respondents from organizations with more than $5 billion in revenue have reached the scaling phase, compared with 29% of those at companies below $100 million.

The story is similar with AI agents, the foundation-model-powered systems that can plan and execute multi-step workflows autonomously. Sixty-two percent of respondents say their organizations are at least experimenting with agents, and 23% report scaling an agentic system somewhere in the enterprise. But that scaling is narrow. In any given business function, no more than 10% of respondents say agents have moved beyond piloting. IT, knowledge management, and marketing are the functions furthest along.

When respondents describe what AI has actually done for their organizations, the answers tend to be qualitative rather than financial. 64% say AI has improved innovation, and roughly 45% report improvements in employee satisfaction, customer satisfaction, and competitive differentiation. Those are meaningful signals, but they sit uneasily alongside the bottom-line numbers: only 39% attribute any level of EBIT impact to AI, and most of that group says it accounts for less than 5% of their organization's earnings.

The picture is more encouraging at the use-case level. In software engineering and manufacturing, 56% of respondents report cost decreases from AI activities in the past year. Marketing and sales leads on the revenue side, with 67% of respondents reporting revenue increases from AI use in that function. The pattern is clear: individual teams are capturing value, but it rarely rolls up to enterprise-wide financial impact.

This disconnect between local wins and company-wide results points to a familiar scaling problem. Organizations have found pockets of AI productivity, but most have not yet redesigned their operating models, data infrastructure, or governance structures enough to compound those pockets into something the CFO can measure.

McKinsey defines AI high performers as the roughly 6% of respondents who report that more than 5% of their organization's EBIT and "significant value" come from AI. These companies look different from the rest in almost every dimension the survey measured.

Their ambitions are bigger. High performers are 3.6 times more likely than others to say their organization intends to use AI for transformative change over the next three years. While 80% of all respondents cite efficiency as an AI objective (and high performers do too, at 84%), the real difference is in the other goals: 82% of high performers also target growth, and 79% target innovation, compared with just 50% for each among the rest.

Their workflows are different. High performers are 2.8 times more likely to have fundamentally redesigned individual workflows around AI. This practice, according to a relative-weights analysis of 31 variables, is one of the strongest contributors to achieving meaningful business impact. They also invest more heavily: more than a third of high performers commit over 20% of their digital budgets to AI technologies, compared with 7% of other respondents. And their leadership shows up. High performers are three times more likely to strongly agree that senior leaders demonstrate real ownership of AI initiatives, including role modeling AI use themselves.

Respondents are divided on what AI will do to headcount. Looking ahead to the next year, 43% expect little or no change in overall workforce size, 32% predict a decrease of 3% or more, and 13% predict an increase of that magnitude. Within individual functions, expectations of workforce reduction are larger than what respondents observed in the past year: a median of 30% expect decreases in the year ahead, compared with 17% who reported them in the past twelve months. Software engineering, IT, and service operations are the functions where the largest shares of respondents anticipate cuts.

At the same time, organizations continue to hire for AI-related roles. Software engineers and data engineers are the most sought-after. Larger companies (those with $1 billion or more in revenue) are roughly twice as likely as smaller ones to report AI-related hiring across nearly every technical role.

On risk, organizations are getting more serious, though not fast enough. Half of respondents from AI-using organizations say they have experienced at least one negative consequence, with inaccuracy the most common at 30%. Inaccuracy is also the risk most organizations are working to mitigate (54%), followed by cybersecurity (51%). The second-most-commonly-experienced risk, explainability, is notably under-addressed: 14% of respondents report consequences from it, but only 28% say their organizations are mitigating it. On average, organizations now mitigate four AI-related risks, up from two in 2022. High performers, who deploy AI in more sensitive contexts, are more likely to encounter problems and more likely to take steps to manage them.