Walk into almost any large organization today and you will find AI everywhere and nowhere at once. Everywhere, in that nearly every function is running a pilot. Nowhere, in that few of those pilots have changed how the business actually operates. The gap between experimentation and impact has become the defining management problem of the moment, and the latest research puts hard numbers on it.

McKinsey's The State of AI in 2025, based on responses from 1,993 organizations across roughly 105 countries, reports that 88% of organizations now use AI in at least one business function, up from 78% a year earlier. Yet only about one third report scaling AI across the enterprise. The remaining majority are stuck in what observers have started calling "pilot purgatory."¹

The leaders did not ask what AI can do. They asked what decision they were trying to improve, and worked back from there.

The same survey found that only a small minority, roughly 6%, see significant enterprise-wide financial impact from AI, while high performers are nearly three times more likely than their peers to have fundamentally redesigned workflows as part of their AI efforts.¹ The lesson is consistent: value comes from changing how work happens, not from bolting a model onto an unchanged process.

Why the gap persists

The blockers are remarkably consistent across industries and company sizes. Data that is siloed and poorly governed cannot feed an enterprise-scale system. Workflows designed for human-only execution produce marginal gains when AI is added without redesign. Functional silos and unclear ownership prevent deployment across the organization. And without KPIs tied to business outcomes, pilots cannot make the case for the next round of investment.

• Anchor to a decision. Every initiative should improve a specific, named business decision. Design the data, the model, and the workflow backward from that decision.
• Keep humans accountable. Governance separates leaders from laggards. McKinsey notes that 51% of firms report AI incidents, and high performers manage that risk with human-in-the-loop rules and clear executive ownership.¹
• Change the operating model. A pilot that does not change how people work produces a demo, not value.

The cost of standing still

The pressure to get this right is rising. In PwC's 29th Global CEO Survey, drawn from 4,454 chief executives across 95 countries, confidence in revenue growth over the next 12 months fell to 30%, down from 38% a year earlier and a recent peak of 56%, with leaders citing uneven AI returns among the reasons.² At the same time, PwC research shows that companies that reinvent well, adapting both their business and operating models, achieve a typical 71% performance premium, a combined measure of profit margin and revenue growth.³

Where to start

Begin with one decision that matters, where better, faster judgment would change the economics. Prove the value end to end, including the change to how people work. Then scale the pattern, not just the model. As Deloitte and ServiceNow put it in their 2026 Workflow Automation Outlook, transformation should be treated not as a project with an end date but as a living discipline embedded across the enterprise.⁴

The organizations that compound advantage are the ones that learn to do this repeatedly, while their competitors are still admiring their pilots.