Recent industry discussions have highlighted how quickly organisations are moving toward agentic AI and how unprepared most remain for its risk implications.
This shift introduces a fundamentally different operating reality. We are moving toward a world where AI systems do not just assist - they execute; and where work is carried out across multiple interconnected systems without constant human intervention. We also foresee a scenario where enterprises may soon manage large populations of non-human digital workers.
This is not an incremental change. It is a step-change in how work is performed, how value is created, and how risk must be managed.
From applications to autonomous actors, traditional enterprise technology is following a predictable model. Users interact with applications, systems process define transactions and controls are applied at known boundaries.
Agentic AI breaks this model. AI agents act on behalf of users or organisations, interact dynamically across systems (CRM, ERP, APIs, cloud platforms), and adapt behaviour based on context and objectives.
To deliver value, these agents require:
· Broad system access
· Ability to trigger actions independently
· Continuous interaction with enterprise data and workflows
This creates a powerful capability but also introduces a new category of operational and security exposure.
Why AI Agents are gaining momentum
The acceleration of agentic AI is not happening because organisations are chasing technology for its own sake. It is happening because AI agents address a practical business need: the ability to automate complex, multi-step work that previously required human coordination across multiple systems.
Potential use cases are already emerging across customer service, IT operations, sales support, finance workflows, knowledge management, and enterprise productivity. In each case, the value comes from giving AI systems the ability to understand context, take action, and coordinate work across applications.
Agentic AI should be viewed as more than a security concern. It is an operating model shift. The risk emerges because the same capabilities that make AI agents valuable—autonomy, access, speed, and scale also make them harder to govern.
Why AI agents change the risk equation
The business case for AI agents is compelling, but it changes the enterprise risk equation. As agents become more autonomous and more deeply connected to systems, organisations must rethink how they govern access, action, and accountability.
1. Access without boundaries
A single agent may connect across finance, customer, operations, and cloud systems—with permissions expanded to enable end-to-end task execution. If compromised, an AI agent does not behave like a traditional endpoint. It becomes a high-privilege, multi-system actor with an expanded blast radius and faster propagation of impact across systems.
2. Autonomous, high-speed actions
Unlike human users, AI agents operate continuously, execute tasks at machine speed, and can chain multiple actions together. This removes natural friction in enterprise processes—meaning risks can materialise and escalate faster than human response cycles, and incidents can propagate before detection or intervention.
3. Dynamic behaviour and intent
AI agents are not static. Their behaviour can change depending on inputs, evolve based on context, and adapt over time. This introduces a critical challenge:
Traditional controls validate identity but agentic systems require validation of intent. It is no longer sufficient to ask who is accessing the system. Organisations must also answer: What is the agent trying to do right now, and has its behaviour deviated from expected patterns?
This shift toward intent and behaviour-based control is still immature in most enterprises today.
4. Explosion of non-human identities
As agent adoption scales, organisations will need to manage large populations of AI agents each with distinct roles, permissions, and behavioural patterns. These non-human identities carry system access, decision-making capability, and autonomous execution rights, while creating limited visibility, unclear ownership, and difficulty enforcing consistent policies.
5. Infrastructure and control models are yet to be ready
Existing enterprise architectures were designed for predictable workloads, human-driven interactions, and segmented control domains. Agentic AI introduces continuous real-time activity, highly dynamic system interactions, and tight interdependence across networking, security, and monitoring.
Legacy control models, however, are struggling to keep pace. This creates structural exposure - not just operational risk.
A growing urgency for organisations
The emergence of AI agents is happening now. Organisations are already experimenting with AI agents in customer service, IT, and operations - embedding agent-driven workflows into core processes and scaling adoption across business units.
However, governance and risk frameworks have not evolved at the same pace. There is a widening gap between - what AI agents are capable of doing and what organisations are able to control and monitor.
As agent adoption accelerates, this gap will expand further-driven by increasing autonomy, greater system integration, and higher volumes of machine-driven activity.
Balancing value creation and risk
Agentic AI introduces a structural shift in enterprise value creation and risk, driven by three converging forces:
· Autonomy – Systems act independently
· Access – Agents operate across multiple systems
· Scale – Non-human identities grow rapidly
Together, these forces create a risk environment that is more distributed, more dynamic, and significantly faster-moving than anything enterprises have managed before.
This is not simply an extension of existing cybersecurity challenges. It is a new frontier requiring a fundamental rethink of governance, control, enterprise architecture, and workforce readiness.
AI agents represent one of the most transformative shifts in enterprise technology - but also one of the most complex. They have the potential to unlock new levels of productivity, enable new operating models, and accelerate business outcomes. But without corresponding control frameworks, they also introduce expanded risk exposure, reduced visibility, and faster, harder-to-contain incidents.
- Kenny Yeo, Director of the Asia Pacific ICT and Cyber Security Practice, Asia Pacific, Frost & Sullivan.
The rise of agentic AI is not just about opportunity, and it is not just about risk - it is about building the operating model to manage both at scale.
Organisations that act early will be better positioned to capture the benefits of agentic AI while managing the risks of an increasingly autonomous digital workforce.
We will continue to examine this evolution across multiple dimensions - from enterprise value creation and deployment models to practical technology choices, governance requirements, risk management, and security implications as organisations move from experimentation to real-world agentic AI adoption.
Kenny Yeo currently leads Frost & Sullivan’s ICT practice across the Asia Pacific.
