IBM Delivers Agentic AI to Networking

IBM Delivers Agentic AI to Networking

Today, IBM is unveiling IBM Network Intelligence, a network-native AI solution that addresses the escalating complexity of modern telecommunications and enterprise networks. Developed in...

Today, IBM is unveiling IBM Network Intelligence, a network-native AI solution that addresses the escalating complexity of modern telecommunications and enterprise networks. Developed in collaboration with IBM Research, the solution is essential for organizations that want to transform every phase of network operations and build trustworthy AI.

IBM Network Intelligence pairs advanced time-series foundation models with LLM-powered reasoning agents to create a network-aware AI collaborator. We believe this approach is critical to addressing the complexity of modern networks where network teams struggle to manage through tools and manual processes; these complex issues represent only a subset of total issues, however, and consume the vast majority of a network team's effort. It is in both the time and resources spent reacting to, understanding, and resolving critical issues where there is untapped value. This reactivity can be a major barrier to accomplishing scalable automation.

Cross-Domain Network Data: A Context Crisis

The core challenge is that network data, ever growing and flowing, is fundamentally defined by relationships and connections scattered across an organization. Beyond just volume, fragmentation across expansive network domains, vendors, and formats creates isolated data system silos that obscure critical insights. But today's tools can't capture or analyze these connections effectively and lack the ability to understand real-time behaviors and cross-domain relationships.

As a result, humans are responsible for doing what the tools can't, stitching together siloed data insights for cross-domain analysis. It can be a slow, labor intensive, and error-prone process that loses this essential network context and one that may hold back important resources from more value-add activities.

While advances in network tools, automation platforms, and traditional ML have helped with many pain points, human operators still can't scale to meet the growing demands of real-time service assurance, low-latency performance, and zero-downtime expectations.

Trusted AI, Delivered via a Dual Intelligence (Analytical and Reasoning) Approach

IBM Network Intelligence is built to meet these challenges head-on by introducing a human-AI partnership model built on complimentary deep learning and agentic technologies. It combines analytical AI that consumes and understands massive amounts of network data, with reasoning and agentic AI that understands that data and the relationship of various inputs across systems, hypothesizes potential issues and root causes, and assists with filtering out noise and delivering high-confidence insights. Together, they are designed to enable a powerful new operating model where AI handles scale, pattern recognition and goal-oriented action, and humans guide the context, judgment, and trust-building required to make that intelligence reliably actionable.

This combination of both analytical and "reasoning" AI capabilities represents a transformational opportunity to embed advanced domain-aware AI across the network. It is intended to improve collaboration between humans and AI agents. Such collaboration sets the path for organizations to stop battling for point automation and start achieving a virtuous cycle where networks evolve continuously, resilience improves, and operations scale effortlesslywithout tool bloat or being swamped with technical debt.

"Analytical Intelligence:" Analytical AI for Network Telemetry and Data Intelligence

The analytical AI capabilities in IBM Network Intelligence are underpinned by IBM Granite Time Series Foundation Models, compact AI models developed by IBM Research as part of the trusted, open IBM Granite family.

What's unique about these models is that they are customized and purpose-built for networking, pre-trained on high-volume telemetry, alarms, and flow data across diverse environments. Unlike purely statistical ML, rule-based tools, or generic LLMs, these Time Series Foundation Models offer deep contextual understanding of network behavior. This approach was created to enable more accurate network observations - to find typically alert-less hidden issues and even provide threshold-less early-warning of degradations, which is essential for building trust in an autonomous system, intended to provide an improved signal-to-noise ratio.

This new IBM offering is built to provide one pipeline for all the different types of networking data used by an organization. This includes how their network is designed, the vendor(s) they use, their operational procedures, and any other documented rules or guidance that are specific or relevant to that organization for running their network.

IBM Network Intelligence pre-trained models are intended to provide organizations with value right away.

"Reasoning Intelligence:" Generative AI for Contextual Reasoning and Automation

The "analytical intelligence" part of the dual-architecture approach of IBM Network Intelligence feeds the networking data to the "reasoning intelligence" side.

Generative AI adds contextual reasoning and automation through an agentic framework powered by LLMs and network context. AI agents, built with IBM watsonx technologies, work together to detect issues, identify probable causes, and generate possible remediation plans. These agents guide troubleshooting, support triage, and offer remediation plans, all with the right amount of human-in-the-loop control for explainability.

These agentic AI capabilities enable iterative anomaly detection and explanation, automating root cause analysis across siloed data sources. The capabilities are also engineered to replace manual, multi-team "war room" processes with a continuous, explainable system that surfaces issues other tools might miss. By specializing in time-series patterns and network semantics, you can address false positives and filer out noise, surfacing only high-confidence, actionable insights. A key benefit of this new offering is that it enables organizations to set the pace - gradually adopting agent-driven actions in the journey toward autonomous network operations.

IBM Network Intelligence has the audacious goal of expanding this human-AI partnership model into the design and build phases of the network lifecycle. Building the necessary trust will take time; most operations teams will likely start by deploying the software alongside existing performance and event management systems, to provide a "second opinion." Once this crucial step establishes trust and explainability, customers will be comfortable enough to move beyond "scripted automation" and toward more reliance on network-aware AI tools. This is how we will see the transformational change needed in networking, where barriers to automation are removed, enabling continuous network evolution, enhanced resilience, and effortless operations scaling.

To learn more, join an IBM webinar "AI and the Future of Networking: Market Trends That Matter," on Thursday, October 9 at 9:00 am ET. You can register here. And, for more information about IBM Network Intelligence, please visit https://www.ibm.com/products/network-intelligence.