Three out of four enterprises plan to deploy autonomous AI agents within two years. Only one in five has governance in place for those agents. That gap, pulled from Deloitte’s State of AI in the Enterprise 2026 survey of 3,235 business and IT leaders across 24 countries, might be the single most important number in enterprise AI right now. The report, subtitled “The Untapped Edge,” covers three AI shifts: agentic AI going mainstream, physical AI already further along than most realize, and sovereign AI reshaping how companies pick their vendors.
What 3,235 Enterprise Leaders Actually Report
Deloitte’s survey ran between August and September 2025, collecting responses from board members, C-suite executives, presidents, vice presidents, and directors across six industries and 24 countries. This is not a developer survey or a vendor-funded benchmark. It represents the people writing the checks.
The headline finding: worker access to AI tools rose 50% in 2025. But fewer than 60% of workers who have access actually use AI in their daily workflow. More tools, same habits. Twice as many leaders as last year report “transformative” business impact from AI, yet only 34% say they are truly reimagining their business around it.
That 34% figure puts a hard number on the “pilot purgatory” problem everyone talks about. The number of companies with 40% or more of their AI projects in production is set to double within six months. The shift from experimentation to deployment is real, but the gap between deployment and transformation is wider than most executives admit.
Companies seeing the most success share a common pattern: they start with lower-risk use cases, build governance capabilities alongside deployment, and scale deliberately rather than sprinting to cover every use case at once.
Agentic AI: From 23% to 74% in Two Years
Currently, 23% of companies use agentic AI at least moderately. Within two years, that number is projected to hit 74%, with 23% using it extensively and 5% fully integrating AI agents as a core component of operations. No other AI category in the report shows this kind of growth curve.
The use cases with the highest expected impact are customer support (no surprise given the contact center AI boom), supply chain management, R&D, knowledge management, and cybersecurity. A financial services company profiled in the report is building agentic workflows to automatically capture meeting action items from video conferences, draft follow-up communications, and track whether people actually complete their commitments. An air carrier uses AI agents to handle common passenger transactions like rebooking flights and rerouting bags. A manufacturer lets AI agents support new product development by finding the optimal balance between cost and time-to-market.
These examples reflect the pattern the LangChain State of Agent Engineering survey also found: the agents that work in production solve narrow, well-defined problems rather than attempting general-purpose autonomy.
The Governance Gap That Should Worry Everyone
Here is the number that stands out: nearly three quarters of organizations plan to deploy autonomous agents within two years, but only 21% report having mature governance models for those systems. The risk concerns are concrete:
- 73% cite data privacy and security as their top AI risk
- 50% flag legal, intellectual property, and regulatory compliance
- 46% worry about governance capabilities and oversight
- 46% cite model quality, consistency, and explainability
Agentic AI fundamentally changes the governance equation. Traditional AI systems recommend actions for humans to execute. AI agents take actions directly. That requires new infrastructure: clear boundaries for agent autonomy, real-time monitoring of agent behavior, and audit trails that capture the full chain of agent decisions and actions.
The companies getting this right are not waiting until after deployment to build governance. They treat governance as a prerequisite, not a follow-up project. That is a harder sell internally, because governance does not produce demos, but the report’s data suggests it is the differentiator between companies that scale AI successfully and those that stall.
Physical AI: Already Mainstream at 58% Adoption
Physical AI, the category that covers robotics, digital twins, IoT integration, and intelligent monitoring, turns out to be further along than the agentic AI conversation suggests. 58% of companies already use physical AI, and adoption is projected to hit 80% within two years.
The regional breakdown is striking. In Asia-Pacific, 71% of organizations already use physical AI, compared to 56% in both the Americas and EMEA. APAC adoption is expected to reach 90% within two years, driven heavily by manufacturing and logistics operations in Japan, South Korea, and Singapore.
The use cases with the biggest projected long-term impact:
- Intelligent security systems and smart monitoring (21%)
- Collaborative robotics (20%)
- Digital twins (19%)
This aligns with what we covered in our deep dive on physical AI agents and robotics: the convergence of foundation models with physical systems is happening faster in Asia because the manufacturing density creates more immediate ROI.
The misconception is that physical AI is a “future” category. The Deloitte data says otherwise. More than half of enterprises are already there.
Sovereign AI: 77% Now Factor Country of Origin Into AI Vendors
This might be the most underreported finding in the entire survey. 77% of companies now consider country of origin when selecting AI vendors. Nearly three in five companies build their AI stacks primarily with local vendors. And 83% view sovereign AI as important to their strategic planning.
Sovereign AI is the idea that a country, and the companies within it, should deploy AI under their own laws, infrastructure, and data governance frameworks. It is driven by a mix of regulatory compliance (the EU AI Act being the most obvious example), data residency requirements, supply chain security concerns, and straightforward geopolitics.
The regional split tells the real story. Only 11% of companies in the Americas rely on foreign-sourced solutions for the majority of their AI stack. In EMEA, that number is 32%. European companies are far more dependent on non-local AI vendors, which creates a strategic vulnerability that regulators and policymakers are actively trying to close.
Deloitte projects that more than $100 billion will be committed to building sovereign AI compute infrastructure in 2026 alone. That money funds domestic data centers, local model training capacity, and regional cloud infrastructure designed to keep data within national borders. For companies selling AI tools and services, sovereign AI is not a policy talking point; it is a procurement filter that 77% of their potential customers already apply.
The Untapped Edge: Why Execution Lags Adoption
The report’s subtitle, “The Untapped Edge,” captures the core tension. AI adoption is accelerating. Tool access is expanding. But the gap between having AI and using AI to transform the business keeps widening.
The pattern Deloitte identifies across all three AI categories (agentic, physical, sovereign) is the same: deployment outpaces the organizational capability to extract value from it. Companies buy the tools, launch the pilots, and then stall when they need to change processes, retrain teams, and build governance structures.
The numbers paint this clearly:
- Worker access up 50%, but daily usage below 60%
- Twice as many leaders report transformative impact, but only 34% reimagine the business
- 74% plan agentic AI within two years, but 21% have governance
For teams planning their AI strategy, the Deloitte data points to a specific action: stop optimizing for more AI deployments and start optimizing for extraction of value from existing ones. The edge is not in adopting one more tool. It is in closing the gap between the tools you have and the outcomes you need.
Frequently Asked Questions
What percentage of companies use agentic AI according to Deloitte 2026?
According to Deloitte’s State of AI in the Enterprise 2026 report, 23% of companies currently use agentic AI at least moderately. This is projected to rise to 74% within two years, with 23% using it extensively and 5% fully integrating it as a core operational component.
What is the AI governance gap identified in the Deloitte report?
Nearly three quarters of organizations plan to deploy autonomous AI agents within two years, but only 21% report having mature governance models for those systems. 73% cite data privacy and security as their top concern, followed by legal and regulatory compliance at 50%.
How widespread is physical AI adoption in enterprises?
58% of companies already use physical AI, with adoption projected to reach 80% within two years. Asia-Pacific leads at 71% adoption, compared to 56% in both the Americas and EMEA. Key use cases include intelligent security systems, collaborative robotics, and digital twins.
What is sovereign AI and why does it matter for enterprises?
Sovereign AI means deploying AI under a country’s own laws, infrastructure, and data governance frameworks. 77% of companies now factor country of origin into AI vendor selection, and 83% view sovereign AI as strategically important. Over $100 billion is expected to be committed to sovereign AI compute infrastructure in 2026.
How many leaders were surveyed in the Deloitte State of AI 2026 report?
Deloitte surveyed 3,235 business and IT leaders across 24 countries and six industries between August and September 2025. Respondents included board members, C-suite executives, presidents, vice presidents, and directors.