Worker access to AI tools jumped 50% in 2025, according to Deloitte's latest State of AI in the Enterprise report. But here's the uncomfortable number buried in the data: 37% of organizations are still using AI at surface level with minimal changes to how they actually work.
That gap between access and real adoption tells the story of where enterprise AI actually stands right now.
The Three Tiers of Enterprise AI
Deloitte's survey splits companies into three groups. At the top, 34% are going deep, building new products or fundamentally reinventing processes around AI. In the middle, 30% are redesigning key workflows. And at the bottom, that 37% majority is essentially bolting AI onto existing processes and calling it a day.
The productivity numbers reflect this split. Two-thirds of companies (66%) report efficiency gains from AI, and 53% say they're making better decisions. But only 20% have actually grown revenue through AI, even though 74% say revenue growth is the goal. That 54-point gap between aspiration and reality is where most enterprise AI strategies are quietly failing.
The Governance Problem With AI Agents
The report flags a timing issue that should worry anyone deploying AI agents (autonomous AI systems that take actions on your behalf, like scheduling meetings, processing invoices, or managing customer requests). Only 20% of companies have mature governance models for these agents, yet adoption is accelerating.
This matters because AI agents operate with less human oversight than a chatbot. When a chatbot gives bad advice, a human can catch it before acting. When an agent acts on its own, the mistake is already made. Running agents without governance is like giving a new employee full system access on day one with no training and no supervisor.
Physical AI Is the Quiet Story
58% of companies report at least limited use of physical AI (think robotics, computer vision on factory floors, autonomous vehicles in warehouses) and Deloitte projects that will hit 80% within two years. Asia Pacific is leading early implementation.
The survey also surfaces what Deloitte calls a "preparedness paradox." 42% of leaders think their AI strategy is solid, but far fewer feel confident about their infrastructure, data quality, risk management, or talent pipeline. The skills gap remains the single biggest barrier to integration, and most companies are prioritizing education over actually redesigning roles or workflows.
That last point is telling. Training people to use AI tools is the easy part. Rebuilding your operations so AI fundamentally changes how work gets done? That's where the 37% surface-level adopters are stuck, and it's where the real competitive gap will open up over the next year.