IBM has released an autonomous network solution designed for telecoms and enterprise networks.
IBM Network Intelligence, developed in collaboration with IBM Research, combines AI agents with specialized time-series foundation models (TSFMs), large, pre-trained AI models that learn generalized patterns from massive amounts of diverse time-stamped data.
“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,” IBM Software Networking’s Benjamin Hickey wrote in a company blog post.
IBM Network Intelligence’s analytical AI capabilities are powered by the vendor’s Granite Time series of AI models. These compact models created by IBM Research are designed to identify hidden issues that typically trigger no alerts and offer early warnings of potential degradations that don’t rely on predefined limits.
Analytics from the TSFMs are then filtered through agentic and reasoning AI, which interprets data and the relationships between inputs across network systems. This can help network teams identify potential issues and root causes, whilst filtering out network noise to provide more assured insights.
IBM gave one telecom and enterprise use case centering around Multiprotocol Label Switching (IP/MPLS) silent drops, where packets are discarded or lost without generating an explicit error message or alert.
The vendor claims agentic AI can spot issues missed by other software tools, while tracing root causes across network layers and accelerating remediation. For network teams, IBM argues, this can provide clear visibility and a drop in escalations without parsing event logs and switching through various tools.
Another use case concerning radio access network (RAN) domain optimization posits the detection of congestion, interference, and imbalances before customers are affected, with agents fine-tuning networks in near real time.
“[IBM Network Intelligence] is built to provide one pipeline for all the different types of networking data used by an organization,” said Hickey. “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.”
Agents afoot
IBM’s solution comes as agentic AI increases its foothold in the networking space. Examples include Extreme Networks, which released an agentic AI-focused network visualization offering, while generative AI Intel spinout Articul8 unveiled a Network Topology Agent capable of creating a queryable graph of entire networks.
Hewlett Packard Enterprise (HPE) also added agentic AI management capabilities to its GreenLake platform this year to provide root-cause analysis for network issues.
On the telecom side, O2 Telefónica Germany recently launched an agentic AI solution in partnership with Tech Mahindra and Nvidia, echoing a Nokia AI agent suite from this summer designed to help telecom operators manage their networks, and an artificial intelligence for IT operations (AIOps) platform from Ribbon Communications.
Like the Ribbon release, IBM is gearing service providers and enterprises to transition to autonomous networks. With a more holistic approach backed by IBM’s compute backbone, IBM Network Intelligence may present one of the more considerable agentic offerings for network teams so far.