Industry executives and experts share their predictions for 2026. Read them in this 18th annual VMblog.com series exclusive.
By Jamie Hutton, CTO, Quantexa
The phase of shiny prototypes and tightly controlled AI experiments has largely passed. Enterprises are now discovering that agentic systems lacking trusted, contextual data and transparent reasoning do not fail quietly once they are pushed into real-world operations. These systems are increasingly being placed in roles where decisions carry real consequences and little room for error.
That shift is what separates agentic AI from earlier approaches. Unlike previous systems that primarily supported human decision-making, agentic systems can initiate actions, chain decisions together, and adapt continuously using inputs from internal systems and external data sources. That autonomy is precisely what makes them powerful and risky.
Most governance frameworks were built for models that operate in isolation. Agentic AI does not behave that way. Its reasons across domains navigate ambiguity and operate over time. As enterprises move these systems into production, that gap widens quickly and becomes operationally unavoidable.
From bounded models to unbounded decision paths
Many organizations continue to treat agentic AI as an extension of traditional machine learning. Agents are piloted in narrow use cases with limited objectives and confined to carefully selected datasets. This approach works only as long as the system remains disconnected from the complexity of the real enterprise.
To deliver meaningful value, agentic AI must operate with context. This involves drawing on customer data, transactions, operational systems, policies, regulations, and third-party intelligence. Each additional connection improves decision quality, but it also expands the decision surface area where things can go wrong.
Errors in this environment are not benign. A flawed assumption or incomplete view of context can cascade through automated actions before human oversight has a chance to intervene. In sectors such as financial services, public sector, and critical infrastructure, the cost of those failures can be immediate and substantial.
As agentic systems begin to shoulder more responsibility, enterprises are finding that traditional governance models start to fall apart. When decisions are automated and linked together, control must be designed into the system from the start, rather than being layered afterward. Visibility into data sources, reasoning paths, and decision logic becomes essential once autonomy extends beyond isolated use cases and into connected, real-world operations.
The rise of evidence-centric AI
As agentic AI moves pilot programs into production, enterprises are drawing stricter lines around what they are willing to trust. Systems that act with autonomy but offer little visibility into how decisions are made can quickly raise red flags, especially when those decisions begin to affect customers, compliance with posture, or financial outcomes.
Concerns like hallucinations, bias, and autonomous errors are no longer confined to academic debate. They show up as operational risk. Boards and regulators are paying closer attention, and risk teams are increasingly uncomfortable with AI-driven decisions that cannot be tied back to reliable data or explained in plain terms.
This is pushing AI into the same accountability framework that already governs other high-impact decisions inside the enterprise. Financial reporting, compliance processes, and operational controls all depend on auditability and evidence that can stand up to scrutiny. As AI assumes a more direct role in decision-making, these expectations are starting to apply just as firmly.
Where pressure builds first
Sectors where automated decisions have real-world consequences will encounter these constraints first. Financial services, public sector agencies, healthcare organizations, and critical infrastructure providers are already feeling pressure to justify how automated decisions are made.
In these environments, governance is not a constraint to innovation. It is the mechanism that allows innovation to scale responsibly. Clear accountability frameworks provide leaders with the confidence to delegate more authority to autonomous systems without compromising control.
The organizations that emerge as leaders will not be those that adopt agentic AI most aggressively. They will be the ones who can prove that their systems are acting on connected, trusted data and reasoning paths that they are willing to defend.
Agentic AI is changing how intelligence shows up inside the enterprise. The real test is not what these systems are capable of, but whether organizations are prepared to take responsibility for the decisions they allow them to make.
##
ABOUT THE AUTHOR
Jamie Hutton is the Co-founder and Chief Technology Officer of Quantexa, where he leads the company’s global research and development organization in advancing its market-leading Decision Intelligence Platform. With over two decades of experience pioneering data-driven technologies, Jamie has been at the forefront of innovations that connect and unify data at scale to solve complex real-world challenges.
