88% of Companies Use AI. Only 5% Are Making Money from It

88% of organizations now use artificial intelligence regularly in at least one business function. According to McKinsey's State of AI 2025, that figure is up from 78% the year before. It sounds like a transformation. But read the number that follows it: only 5.5% of those companies report that AI has had a significant impact on their EBIT. The adoption race is real. The payoff, for the vast majority of businesses, is not.

For business owners watching competitors announce AI initiatives, the pressure to act feels urgent. But the data tells a different story. The real competitive advantage is not in having AI. It is in deploying it in a way that actually changes how your business generates value. And that distinction is one that most organizations are currently failing to make.

The Adoption Numbers Are Misleading

The 88% adoption figure is striking but hollow as a measure of progress. McKinsey's research shows that only 39% of organizations report any enterprise-wide EBIT impact from AI at all. And within that 39%, most say AI contributes less than 5% of EBIT. The remaining 61% of companies are running AI somewhere in their business and seeing no measurable effect on their bottom line.

This gap has an industry name: "pilot purgatory." Companies launch AI projects, generate internal enthusiasm, run proofs of concept, and never cross the threshold into actual business value. Only about one-third of organizations have managed to scale AI across their enterprise. Two-thirds are stuck between experimentation and transformation, spending real budget without returning real results.

The pressure to announce AI adoption has created a widespread category of initiatives that exist in presentations and press releases but not in the places where money is actually made or saved. A company can have 40 AI projects running simultaneously and still see no meaningful change to its revenue, cost structure, or competitive position. McKinsey's data shows this is the norm, not the exception.

Why AI Projects Are Being Canceled

Gartner's 2025 research adds a sharper edge to this picture. The analyst firm predicts that more than 40% of agentic AI projects, those involving autonomous AI systems capable of completing multi-step tasks without continuous human input, will be canceled by the end of 2027. The causes are consistent: escalating costs, unclear business value, and inadequate risk controls.

Part of the failure pattern comes from what Gartner calls "agent washing." Vendors are rebranding existing chatbots, robotic process automation tools, and basic workflow software as AI agents without making the underlying technology meaningfully more capable. Gartner estimates only around 130 of the thousands of companies claiming to sell agentic AI are building the genuine article. For businesses buying these products, the gap between the sales pitch and real-world performance is a primary driver of project cancellation. The ROI calculation promised during procurement simply never closes.

The execution challenges are also intensifying as organizations move from curious to committed. KPMG's Q4 2025 AI Pulse Survey found that 65% of business leaders now identify agentic system complexity as the main barrier to AI strategy execution. That number has remained stable for two consecutive quarters, suggesting organizations are not adapting fast enough to overcome it. Cybersecurity concerns have risen to 80% of respondents, up from 68% at the start of 2025, and data privacy worries have climbed from 53% to 77% over the same period. These are not the concerns of organizations that are still exploring AI. They are the concerns of organizations deep enough in implementation to understand what the real problems look like.

What the Top 5% Do Differently

McKinsey's data identifies a consistent pattern separating high performers from everyone else. The most important difference is not budget size, not access to better models, and not headcount. It is how companies think about the relationship between AI and their existing operations.

High-performing organizations redesign workflows from scratch rather than layering AI onto processes that were designed before AI existed. This distinction is fundamental. A "tool overlay" approach, adding an AI assistant to an existing process without changing how the process works, produces marginal gains at best. It reduces time on certain tasks, improves output consistency, and generates positive user feedback. But it does not change the underlying logic of the process, so it cannot change the underlying economics. Companies that ask instead, "how would we design this process if AI had existed from the start?", tend to reach entirely different structural answers. Those answers are where the 5% EBIT impact lives.

The second differentiator is measurement discipline. High performers establish specific KPIs connecting AI initiatives to business outcomes before deployment starts. They are not tracking adoption rates, prompt counts, or employee satisfaction with AI tools. They are measuring revenue, cost, and margin impact directly. This forces a harder upfront question: what does success look like in financial terms? And it creates accountability that keeps projects from surviving on enthusiasm alone.

The third factor is investment commitment. More than a third of AI high performers allocate over 20% of their total digital budget to AI, according to McKinsey. These organizations are five times more likely to make a major strategic AI bet than average companies. Budget scale signals organizational conviction, and conviction drives the kind of cross-functional coordination that transformation requires. Companies treating AI as an IT line item rarely achieve the operational restructuring that high performers pursue.

None of these differentiators are technological conclusions. They are operational and organizational ones. The companies generating real returns from AI are not using superior models or accessing technology unavailable to their competitors. They are making different strategic choices about where to aim, how to measure, and how much to commit.

The Practical Implication for Business Owners

The most useful takeaway from this data is also the most uncomfortable one. If you are currently running AI tools on top of existing business processes without redesigning those processes, you are statistically likely in the 95%, not the 5%. That is not a criticism of your AI investments so far. It is a description of where almost every organization sits right now.

Adding an AI writing tool to your marketing workflow, deploying a chatbot on your website, or using AI to summarize documents and emails can reduce friction and save hours. These are real benefits. But they will not produce enterprise-level EBIT impact because they do not touch the underlying structure of how your business operates. The cost of the process remains largely the same. The revenue logic remains unchanged. You have made an existing process slightly faster, not a fundamentally different one.

The question worth asking in 2026 is not "which AI tools should we be using?" but "which of our core business processes would look completely different if we assumed AI could handle the parts we currently staff with humans?" That question leads to different answers, different projects, and different financial outcomes. It is also a harder question, which is why most organizations are not asking it.

Gartner projects that by 2028, at least 15% of day-to-day business decisions will be made autonomously by AI systems, up from essentially zero in 2024. That transition will not happen through incremental tool adoption. It requires businesses willing to rebuild workflows from the foundation up, with a clear financial thesis about what changes and why.

The 88% adoption figure is real. So is the 5.5% impact figure. The gap between them is not a technology problem. Companies on both sides of that gap have access to the same models, the same platforms, and largely the same vendor options. The difference is a strategy problem. And strategy problems have owners.