IT leaders are understandably hopeful about embracing agentic AI, a technology that promises to address unprecedented network complexity while improving network resilience and enterprise security posture.
What is Agentic AI?
According to Gartner, “Agentic AI refers to a class of systems developed using various architectures, design patterns and frameworks, encompassing both single AI agent and multi-agent designs. These systems are capable of performing unsupervised tasks, making decisions and executing end-to-end processes.” As AI agents can automate complex processes, technology executives I have spoken with are hopeful that agents will supercharge operations, reduce human error, proactively resolve performance anomalies and respond to security incidents remotely. Agentic operations could complete tasks that used to take hours, days or weeks, in just a few minutes. Yet, alongside unprecedented efficiency comes uncharted risk.
The Risks of Human Error and Machine-Speed Propagation: Lessons From the CrowdStrike Outage
Seemingly minor errors, such as misconfiguration, continue to be a leading cause of network and IT outages. The July 2024 CrowdStrike outage is a stark reminder of this fact. A faulty sensor update from the Falcon platform resulted in the most significant IT outage in history, impacting over 8.5 million Windows devices worldwide, disrupting flights, shuttering hospitals and public transit, derailing financial institutions and costing U.S. Fortune 500 companies an estimated $5.4 billion in damages.
Agentic AI didn’t cause this incident, but it does demonstrate the potential repercussions of a single mistake. In this case, millions of systems crashed and the recovery was expensive. The impact of a similar, simple yet significant misconfiguration could have unfolded even more rapidly and widely if an autonomous agent had deployed it at machine speed, overwhelming human recovery efforts and resulting in both financial and operational consequences.
This example, along with countless others I’ve heard about, working with Fortune 500 IT leaders clearly demonstrates the need for agents that operate on trustworthy, accurate data and can rely on guardrails to prevent errors.
Understanding the Exact State of the Network
For agentic AI to deliver trustworthy outcomes, it must operate on mathematically precise network data. Before implementing agentic AI, enterprises require detailed, always-current data, including knowledge of all devices on the network, their configurations and state, access control policies and every path that a packet can travel across the network. Data must be presented in a vendor-agnostic manner and structured to preempt ambiguity and minimize errors for agents. Without this foundation, AI agents risk acting on incomplete or outdated information, leading to configuration drift, security policy violations or unexpected service disruptions.
Crucially, this data cannot be generic or extrapolated from similar environments; it must reflect the exact state of the specific network where the agent is operating. Using generalized information introduces significant risks. Configurations, policies and interdependencies can differ significantly between networks (or subnets) based on topologies, policy intent or operational constraints. When agentic AI consumes data that isn’t tailored to the network at hand, or is incomplete, it is much more likely to make incorrect decisions, potentially propagating errors at scale and speed.
Fixing the Foundation: How a Network Digital Twin Delivers Confidence in Agentic Operations
The solution to safe, agentic operations lies in establishing an always-accurate, network-specific single source of truth that only a true network digital twin can provide. A true network digital twin is a mathematically precise virtual copy of an organization’s network, reflecting the current state and configuration of every connected device and cloud instance within the hybrid multi-cloud environment. A true network digital twin understands network policy, segmentation and intended behavior, and can verify compliance, notifying the humans in charge in the event of any variance between desired and actual behavior. This single source of ground truth delivers the precise data foundation that AI agents require to successfully enhance network operations and security.
Opportunity Anchored in Precision
Agentic AI can transform network operations by reducing manual workload and human error, and by enabling strategic innovation. Legacy tools were not built to handle this. Realizing AI’s promise requires agentic operations to be anchored in always-current, mathematically precise data. With the proper foundation in place, enterprises can confidently deploy agentic operations and free up engineering resources to focus on strategic imperatives.
