Unlocking Trust in Agentic AI: The Identity Governance Challenge

In today’s digital landscape, agentic AI is revolutionizing industries, from healthcare to manufacturing. Yet, despite the incredible potential of these intelligent agents, many enterprises are struggling to move beyond pilot projects. The key barrier? Identity governance.

Imagine a scenario where a doctor in a hospital exam room observes a medical transcription agent seamlessly updating electronic health records, suggesting prescription options, and accessing patient history in real time. Similarly, a computer vision agent on a manufacturing line is conducting quality control inspections at a speed unmatched by human inspectors. These agentic AI systems create non-human identities that pose a challenge for enterprises to manage effectively.

Cisco President Jeetu Patel highlighted this issue, stating that while 85% of enterprises are running agent pilots, only 5% have successfully transitioned to production. The trust gap lies in the fundamental questions of identity governance: Which agents have access to sensitive systems, and who is accountable for their actions?

According to research by IANS, most businesses lack the necessary role-based access control for human identities, making it even more challenging to manage agent identities. The 2026 IBM X-Force Threat Intelligence Index reported a significant increase in attacks exploiting vulnerabilities in AI-driven systems, emphasizing the importance of robust identity governance.

The Architectural Trust Gap: A Framework for Secure Deployment

Michael Dickman, SVP and GM of Cisco’s Campus Networking business, emphasized the need for a trust framework that integrates security from the outset, rather than treating it as an afterthought. Dickman’s experience in leading technology companies highlighted the critical role of identity governance in ensuring the secure deployment of agentic AI.

Dickman outlined four key conditions for addressing the trust gap in agentic AI:

1. Secure Delegation: Define permissions for each agent and establish human accountability for their actions.
2. Cultural Readiness: Adapt workflows to accommodate agent-scale processing and leverage behavioral data for informed decision-making.
3. Token Economics: Implement hybrid architectures that combine agentic AI reasoning with traditional deterministic tools for efficient execution.
4. Human Judgment: Recognize the irreplaceable role of human judgment in refining AI output and decision-making.

Dickman’s framework underscores the importance of a holistic approach to identity governance, encompassing not only technical controls but also cultural readiness and human oversight.

Building Trust Through Cross-Domain Visibility

One of the key challenges in managing agentic AI identities is the fragmented nature of enterprise data across different systems and domains. Dickman emphasized the value of network telemetry in providing a comprehensive view of system-to-system communications, enabling cross-domain correlation for enhanced security.

By unifying network, security, and application telemetry into a shared data fabric, organizations can gain valuable insights into agent behaviors and interactions. This integrated approach is essential for effective identity governance and policy enforcement in agentic AI deployments.

Addressing Siloed Data and Permission Sprawl

A common pitfall in agentic AI deployments is the siloed nature of agent data, leading to fragmented views and permission sprawl. Dickman highlighted the need for a unified approach to data sharing and correlation, enabling organizations to derive actionable insights from network telemetry.

By adopting a platform strategy that facilitates data sharing across domains, enterprises can overcome the limitations of fragmented observability tools and application platforms. This integrated approach is crucial for addressing the trust gap in agentic AI deployments.

Trust Gap Assessment: A Roadmap for Secure Deployment

Dickman’s trust gap assessment framework provides a structured approach to evaluating and addressing key challenges in agentic AI deployments. By focusing on agent identity governance, blast radius containment, cross-domain visibility, governance-to-enforcement pipelines, and cultural readiness, organizations can enhance the security and reliability of their AI systems.

Implementing microsegmentation for agent-accessible systems, establishing policy-to-enforcement pipelines, and fostering cultural readiness for agent-scale processing are essential steps in building trust and ensuring the secure deployment of agentic AI.

Five Priorities for Secure Agentic AI Deployments

1. Cross-Functional Alignment: Foster collaboration between line-of-business, IT, and security leaders to define clear expectations for agentic AI.
2. Production-Ready Governance: Strengthen identity and access management processes to accommodate agentic workloads and ensure robust policy enforcement.
3. Platform Approach to Networking: Adopt a platform strategy for networking infrastructure to facilitate data sharing and cross-domain correlation.
4. Hybrid Architectures: Design hybrid architectures that combine agentic AI reasoning with traditional tools for efficient execution and token economics.
5. Bulletproof Use Cases: Prioritize high-value use cases with role-based access control, privileged access management, and microsegmentation from day one to build organizational confidence in agentic AI deployments.

By focusing on these priorities and implementing Dickman’s trust framework, organizations can unlock the potential of agentic AI while ensuring the security and trustworthiness of their deployments. Trust in AI is not a luxury but a necessity for the future of enterprise innovation.