Deloitte’s State of AI in the Enterprise 2026 report confirms what many tech observers have suspected: Artificial intelligence is spreading quickly across organizations, but the hardest part, turning adoption into real business advantage, remains uneven.
Based on responses from more than 3,000 director- to C-suite-level leaders across 24 countries, companies describe a clear shift away from experimentation and toward broader deployment. Sanctioned access to AI tools is now available to roughly 60% of workers, up from under 40% a year ago. That’s a big jump in a short time, and it suggests many firms have moved past internal debates over whether AI is a core strategy.
But access is not the same as real workers using it for real productivity. Deloitte’s research points to underuse even among those who can use approved tools, a clear sign that offering tools doesn’t necessarily equate to changing work habits.
The same pattern appears in the long-running problem of pilot fatigue. Deloitte found that only about a quarter of organizations have moved 40% or more of their AI pilots into production. Many respondents say they expect to get there soon, but the current state is still a bottleneck. There are plenty of proofs of concept, yet fewer scaled systems embedded in real AI workflows.
Optimizing the System, Not Yet Redesigning It
Where AI is making a difference today, the impact is more operational than transformational. Most leaders report productivity and efficiency improvements, and Deloitte says the number of executives who view AI as having a transformative impact on their organization has risen sharply versus last year.
Yet only about a third of companies say they are using AI to deeply transform their business. In other words, many firms are optimizing the system they already have, while fewer are redesigning their systems.
The revenue picture is concerning for those companies that have invested heavily. Deloitte’s survey data suggests that while nearly three-quarters of organizations want AI to drive revenue growth, only about one in five say it has done so to date.
That disconnect is now a central tension in enterprise AI: boards want growth, teams deliver speed-ups, but the bridge between the two is often a messy mix of data challenges and process rebuilds.
Deloitte’s leadership frames the moment as a pivot point. Nitin Mittal, Deloitte’s global AI leader, stresses that companies need to better combine systems and staff. “As organizations look to unlock AI’s full value, leaders should enable enterprise value by consciously weaving AI into the fabric of their business workflows and through the better coupling of people and machine intelligence.”
Agentic Faces Challenges
That focus on operational change matters even more as agentic AI finds great enthusiasm, despite serious logistical challenges with integration. Deloitte’s survey indicates that nearly three-quarters of companies plan to deploy autonomous AI agents within two years, and a large majority expect they will customize agents to fit their own needs.
Yet agent autonomy, for all of its promise, is highly difficult. Deloitte found that only a small minority of organizations say they have mature governance for agents. That governance gap is an enormous problem. An AI agent that can trigger a customer communication or update records can also make errors at machine hyper-speed, unless companies define guardrails and accountability that can be automated.
Two other categories are rising alongside agents. Physical AI, which is AI embedded in real-world systems such as robotics, sensors, and autonomous equipment, is already in use at more than half of surveyed firms, with adoption expected to climb further.
And sovereign AI, which refers to AI systems that are hosted and controlled to meet local laws and strategic concerns, is becoming a procurement factor. Deloitte’s survey suggests companies increasingly care about country of origin and are leaning toward local vendors for key parts of their AI stack.
Big Shifts Take Time
Taken together, the report describes what might be thought of as a source of conflict in the enterprise: pressure to move quickly, paired with the reality that governance and workforce change don’t move at the speed of model releases.
The companies that gain the most competitive advantage will be those that turn tool access into real usage, and shift from automation into new ways of operating, and both those changes now appear if they will require years to fully implement.
