Enterprise AI moves from pilot to production in 2026, but gaps in governance and talent persist

Worker access to AI across enterprises rose 50% in 2025, and the number of companies with at least 40% of their AI experiments running in production is on track to double within six months. Those figures, from Deloitte's 2026 State of AI in the Enterprise report, capture the central tension defining enterprise AI right now: ambition is outrunning execution.

A separate survey by NVIDIA, drawing on more than 3,200 responses across financial services, retail, healthcare, telecommunications, and manufacturing, found that 64% of organizations are already actively using AI in operations. Another 28% are still in the assessment phase. The message from both reports is consistent: the pilot era is ending, and the scaling era has begun.

Productivity leads, revenue follows slowly

So far, efficiency is where enterprises are actually seeing returns. Deloitte found that two-thirds of organizations, 66%, report productivity and efficiency gains from AI. After that, benefits thin out considerably: 53% cite enhanced insights and decision-making, 40% report cost reductions, and just 20% say AI has grown revenue. NVIDIA's survey reinforced the productivity story, with more than half of respondents (53%) naming improved employee productivity as one of AI's biggest operational impacts.

Revenue growth is more aspiration than reality at this point. Deloitte found that 74% of organizations hope to grow revenue through AI in the future, compared to the 20% already doing so. In telecommunications specifically, NVIDIA's report found that 99% of respondents said AI had improved employee productivity, with a quarter describing the improvement as major or significant.

Scaling without redesigning

More access to AI has not automatically produced deeper organizational change. Deloitte's report breaks companies into three groups by how they are using AI: 34% are deeply transforming their business by creating new products, reinventing core processes, or rethinking business models. Another 30% are redesigning key processes. The remaining 37% are using AI at a surface level, with little or no change to existing workflows.

Workforce strategy reflects the same pattern. Deloitte found that education, training employees on AI tools, was the top talent response to AI adoption. Redesigning roles or workflows around AI capabilities ranked lower, and a significant share of companies have not redesigned jobs at all. The AI skills gap remains the most commonly cited barrier to deeper integration, according to Deloitte. NVIDIA's surveys flagged the same issue: the lack of qualified AI experts consistently appeared as the biggest adoption challenge across industries.

Agentic AI surges ahead of its guardrails

The next phase of enterprise AI is already arriving faster than governance frameworks can handle. Deloitte's report found that agentic AI, autonomous systems that act and decide with minimal human oversight, is poised for a sharp rise in enterprise use over the next two years. Today, 23% of companies report at least moderate use of agentic AI. That figure is expected to grow substantially. The problem: only one in five companies currently has a mature governance model for these autonomous agents.

NVIDIA's surveys corroborate the agentic momentum. Across industries, companies reported moving from deploying AI as a static tool to building systems that can take sequences of actions autonomously. The governance gap Deloitte identifies is not a niche compliance concern, it has direct operational risk implications as agentic systems increasingly touch customer interactions, financial processes, and logistics chains.

Physical AI and sovereign strategy enter the picture

Beyond software-based AI, physical AI is expanding quickly. Deloitte reports that 58% of companies have at least limited use of physical AI today, robots, digital twins, intelligent monitoring systems, and autonomous logistics. That number is projected to reach 80% within two years, with the Asia Pacific region leading in early implementation. In manufacturing and logistics specifically, NVIDIA's report noted digital twins as a key productivity tool, with companies using AI-powered simulations to boost efficiency on factory floors.

Geopolitics is also shaping AI strategy. Deloitte introduced the concept of sovereign AI, deploying AI under a country's own laws, infrastructure, and data governance frameworks, as an emerging strategic priority. The report found meaningful shares of executives factoring in an AI solution's country of origin when making vendor decisions, and building their AI stacks primarily with local vendors. For multinationals, the implications for procurement and platform strategy are material.

Large companies pulling further ahead

Company size is a reliable predictor of AI maturity. NVIDIA found that more than three-quarters (76%) of respondents from companies with over 1,000 employees report active AI usage, versus much lower rates at smaller firms. Larger organizations have greater capital for infrastructure and data science talent, which allows them to push projects from pilot to production on more specific, higher-impact use cases, and to report greater ROI.

Deloitte's preparedness data adds nuance to that picture. While 42% of companies now say their AI strategy is highly prepared, up from last year, confidence drops sharply when the question turns to infrastructure, data management, risk and governance, and talent. The gap between strategic confidence and operational readiness may prove to be the defining challenge as enterprises attempt to move from isolated AI wins to enterprise-wide transformation. Deloitte's next reporting cycle will track whether the expected surge in production deployments materializes on schedule.