What Agentic Systems Really Mean for Organisations
Jan Esman
Jeroen Bolluijt
This article draws on insights from Inside AI: The Agent Problem Nobody’s Ready For, a CTO Consulting podcast featuring Jan Esman and Jeroen Bolluijt.
The word “agentic” is everywhere right now. Like “AI” three years ago, it has become a marketing term applied so broadly that it risks losing meaning. Beneath the hype, however, something genuinely significant is happening, and organisations that understand the shift early will be better placed to navigate what comes next.
In a recent episode of CTO Consulting’s Inside AI podcast, Jan Esman, Head of Strategy at CTO Consulting, sat down with Jeroen Bolluijt, Lead AI Specialist, to unpack what agentic AI means in practical terms, and why it demands a fundamentally different approach to technology governance.
Beyond Chat: What Makes AI Truly Agentic
Most organisations are familiar with the chat-interface model. You prompt, the system responds, and you act on the result. Agentic AI is different. The first distinction is action. An agentic system does not simply reply. It executes. The more consequential shift, however, is autonomy. When you give an AI system the ability to make decisions, rather than carry out instructions, you give it decision rights. Once groups of agents can interact with one another, spawn new agents, and reconfigure themselves, you are operating in territory that looks far less like software and far more like organisational design.
Think Organisation Design, Not IT Delivery
This is one of the central insights from the conversation. The questions that arise in a multi-agent environment, including who leads, who checks the work, how errors are surfaced, and how quality is maintained, are the same questions that arise when managing teams of people. Agents, like people, can follow unproductive paths. They can obscure errors. They can also seek approval, which means they are not always transparent about failure states they would rather not report.
The practical implication is that governing agentic systems is less a technical problem than a cultural and organisational one. Culture, defined not as something nebulous but as consistent, observable behaviour, becomes the operating framework. If agents behave in ways that are inconsistent with your brand, your values, or your service promise, you have a trust problem, regardless of how well the underlying technology performs.
The Governance Gap
Current frameworks are struggling to keep pace. Singapore’s government released an agentic AI framework in early 2025, accompanied by a 109-page trust guide covering what to test for and how to do it. Detailed as it is, the framework primarily applies to the AI-assisted, human-level stage of agentic deployment. For genuinely autonomous, cross-boundary agent systems, the supporting frameworks barely exist yet.
Two gaps stand out. The first is system governance, which concerns the systems that make decisions, rather than just the decisions themselves. The second is cross-organisational accountability. When agents from multiple departments or organisations interact and make decisions together, who is responsible for the algorithm that coordinates them? These are not questions current governance models were designed to answer.
Interoperability Is the Next Frontier
As organisations move from contained AI experiments to connected agent ecosystems, a new layer of infrastructure becomes necessary. An enterprise AI control layer, vendor-agnostic and capable of managing interoperability across platforms, tools, and organisational boundaries, is emerging as a critical next capability. Several well-funded companies in the United States are beginning to address this need, but no dominant approach has yet emerged.
Trust Is a Sociological Problem
One of the more provocative observations in the conversation is that trust in AI is not primarily a technical challenge. People tolerate human error, but they are far less tolerant of machine error. The bar for autonomous systems is higher, not necessarily because the technology is less reliable, but because of how humans perceive and assign risk. Building trust requires either dramatically reducing perceived risk or delivering benefits so substantial that the risk calculus shifts. That is a challenge in communication, culture, and experience design as much as it is in engineering.
The Anticipation Opportunity
The longer-term promise of agentic AI, particularly in government, is a shift from reactive service delivery to anticipatory service. Rather than waiting for a citizen to request support, an agentic system could identify upcoming needs based on life events, and proactively assemble the right response. This kind of dynamic, responsive service delivery has long been an aspiration. Agentic AI may be what finally makes it achievable.
The conversation is only beginning. Organisations that engage with it seriously now will be far better positioned when the frameworks, infrastructure, and expectations catch up.
About the Contributors
Jan Esman
Jan Esman is a seasoned digital advisory leader with extensive experience in IT strategy, enterprise architecture, and digital transformation. His expertise in aligning technology solutions with business objectives ensures that CTO Consulting clients benefit from strategic insights and effective digital initiatives.
Jeroen Bolluijt
Jeroen Bolluijt is an experienced AI and innovation leader with deep expertise in AI strategy, digital transformation, and emerging technologies. He focuses on translating advanced technologies into practical outcomes by aligning innovation initiatives with organisational priorities, governance, and delivery capabilities, enabling clients to realise measurable value from AI and digital initiatives.