… %!s(
In my conversations with CIOs, chief AI officers and enterprise data leaders over the past year, a theme that has continued to surface is that many organizations are adopting agentic AI faster than they are building the controls required to govern it responsibly.
According to Deloitte's 2026 “State of AI in the Enterprise” report, “Today, 23% of companies are using agentic AI at least moderately. However, within the next two years agentic AI is expected to become nearly ubiquitous, with nearly 3 in 4 companies (74%) using it at least moderately.” But it also uncovered that just “1 in 5 (21%) report having a mature model for governance of autonomous agents.”
What stands out to me is how quickly AI agents are moving beyond being used in experimentation environments and as productivity assistants. I’ve seen them start to orchestrate workflows, interact across enterprise systems, trigger actions and make operational decisions with increasing autonomy.
That changes the stakes dramatically. A flawed recommendation can be reviewed. A flawed autonomous action can become an operational incident within seconds.
How Autonomous AI Is Reshaping Operational Risk At Enterprises
Traditional enterprise applications execute deterministic workflows, whereas autonomous systems operate across constantly changing data, workflows and systems at machine speed.
Based on my discussions with enterprise leaders, many teams are still applying governance models designed for static software environments to systems that continuously learn, adapt and act autonomously. That mismatch is creating a category of operational risk that traditional governance models were never designed to manage.
Managing isolated software errors is relatively straightforward. The larger challenge is that because autonomous systems can now propagate decisions across interconnected workflows, infrastructure and operations in real time, small failures no longer remain isolated. They can cascade operationally before human teams fully understand what happened. The potential consequences of this can be significant, especially in certain industries, such as cybersecurity and financial services.
For example, in cybersecurity, autonomous systems can help teams detect and respond to threats more quickly. But if a false positive automatically isolates production infrastructure or blocks legitimate users, it can rapidly disrupt an organization’s critical operations. In financial services, teams can use AI agents for fraud analysis and compliance reviews. But if leaders can’t clearly explain why a transaction was flagged or how decisions propagated across systems, their organizations can face regulatory exposure and reputational damage.
What ties these examples together is the importance of operational trust. Enterprises struggle to operationalize systems they can’t explain, audit or control.
The Rise Of Operational Trust
A key finding from a 2026 McKinsey report was that “Inaccuracy and cybersecurity remain the most frequently cited AI risks as adoption expands.” One of the biggest misconceptions I’ve seen leaders have about enterprise AI is the belief that access to enterprise data alone creates trustworthy automation. It does not—autonomous systems require business context.
Business context is a continuously maintained understanding of how people, assets, policies, processes and business events connect across the enterprise. It captures the relationships, business rules and current operational state that give enterprise business data meaning.
Without that connected understanding, AI agents risk making decisions operating based on incomplete or disconnected information, rather than the full context of the business, limiting explainability and increasing the likelihood of unintended consequences.
Shared understanding enables autonomous systems to ground decisions in relationships, evidence and business meaning instead of disconnected fragments of enterprise information.
For example, say a clinical research organization analyst is trying to identify sites for a new clinical trial. If the organization uses an AI agent to help with the identification process, it shouldn’t simply sift through a set of documents in one system and retrieve a list of sites that have conducted similar trials. The data needed to provide an accurate recommendation is spread across systems and different data types. It needs to take into account contextual factors such as how enrollment history, patient populations and prior trial performance relate to one another.
From my observations, the organizations making the fastest progress are treating business context as shared enterprise infrastructure rather than rebuilding it for every AI application. They build it once, allowing every agent to operate from the same trusted foundation.
Meanwhile, enterprise AI is moving beyond isolated experimentation. The Deloitte report I referenced earlier found that 30% of the companies it surveyed “are redesigning key processes around AI but keeping their business models intact.” In my view, this is where the conversation around enterprise AI is beginning to shift fundamentally.
A year ago, most of my discussions with executives focused on productivity gains and AI experimentation. Increasingly, those discussions are shifting toward operational trust. As AI becomes more deeply embedded in enterprise operations, I believe that the competitive advantage will belong to organizations that can scale autonomous AI without sacrificing trust, governance and control over autonomous decisions.
Practical Steps Leaders Should Take Now
Based on what I’m seeing across the market, leaders preparing their organizations for successful autonomous AI implementations are focusing on several practical areas early on.
First, they are establishing clear governance and accountability for autonomous systems so AI decisions can be explained, audited and controlled as they become deeply embedded in enterprise operations.
Additionally, they are prioritizing business context before automation by ensuring AI agents can reason from a connected understanding of the business. This gives every AI agent the same trusted foundation for making decisions rather than forcing each one to assemble its own view from isolated systems and records.
Third, they are strengthening explainability and traceability. Specifically, they are gaining visibility into why systems acted in a certain manner, which systems were affected and how decisions propagated across the organization.
Finally, the most effective organizations are starting with high-value, lower-risk workflows to build confidence, refine governance models and establish operational trust. That puts them in a better position to expand autonomous decision-making across the business.
In my view, the enterprises that stand to win with autonomous AI will be those that build the operational foundations that enable every autonomous system to reason from a single, connected and trusted understanding of the business. This understanding is what enables organizations to scale autonomous AI while maintaining explainability, governance and operational trust.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
