Building a unified assurance fabric for cross-domain network operations

Commercial context

Modern networks depend on a vast chain of operational systems that have grown in complexity over many years. Every layer generates its own telemetry, topology, and service data. Each uses different formats, standards, and interfaces. As a result, CSPs can struggle to maintain a single, reliable view of what is happening across the network. Faults that span domains remain difficult to detect, and incidents that cross CSP boundaries are even harder to diagnose. The gap between what CSPs know and what they need to know continues to widen.

This kind of fragmentation is now a critical obstacle. Inventory systems, service models, alarms, and topology engines do not move in sync. A fault that appears in one domain may not register with the tools monitoring another. Key agents operate with incomplete or conflicting context. In a market where premium enterprise services depend on tight SLA guarantees, these inconsistencies have real commercial impact. They increase mean time to repair, create confusion during coordinated response, and fuel unnecessary escalations. They also reduce CSPs’ ability to commercialize differentiated quality-of-service tiers or to support cross-border, end-to-end offers with confidence.

Manual operations compound the issue. Engineers still sift through tickets, logs, and telemetry from multiple systems. They correlate faults by experience rather than by shared logic. This slows resolution and introduces human error. It also makes it difficult for operators to scale new services without adding more operational effort. As networks become more dynamic and distributed, the cost of this model becomes unsustainable. Operators need a way to align network knowledge, automate cross-domain actions, and guide remediation using consistent business intent.

The solution

The PRISM AI Catalyst addresses this challenge by creating a unified, AI-driven assurance fabric designed to operate across domains and across CSPs. The framework starts by synchronizing inventory, topology, and service data in real time. Instead of each monitoring system maintaining its own partial view, PRISM AI uses TM Forum-aligned APIs and data models to generate a consistent context for all participating agents. This eliminates information drift and ensures that observations from one domain can be interpreted correctly by another.

The solution is an agentic architecture powered by the model context protocol (MCP). Autonomous agents use this shared context to interpret events, exchange reasoning, and decide how to act. When a fault appears, they correlate signals against the unified topology rather than isolated system views. They share early hypotheses, adjust their understanding as new evidence arrives, and escalate only when cross-checking indicates a genuine issue. This reduces false positives and helps teams focus on the incidents that matter.

Intent management provides the guidelines for how these agents behave. CSPs define high-level objectives such as restoring service within a defined timeframe, maintaining experience tiers for key customers, or minimizing cross-domain impact. The agents use these objectives to guide remediation. Instead of following fixed playbooks, they adapt actions to the live environment while staying aligned with business goals. On the whole, this moves the industry closer to the TM Forum Autonomous Networks Mission, where closed-loop systems interpret intent and act with precise, consistent logic.

The architecture is rooted in the ODA framework, which ensures modularity and interoperability. Core TM Forum assets used in the Catalyst include TMF921, TMF641, TMF633, TMF638, TMF724, TMF655, TMF639, TMF652, and multiple TMFC components. Together, these assets provide standardized ways to model services, expose operational data, and manage cross-system interactions. This creates a stable foundation for autonomous agents and reduces the cost of scaling the solution across partners, regions, and service categories.

By removing silos and supporting shared reasoning, the framework accelerates end-to-end fault correlation. Agents detect root causes faster because they see the entire context, not a fragment. The integrated loops reduce false incidents and drive down mean time to repair by up to 50 percent. Faster resolution improves customer experience and strengthens operators’ ability to meet strict SLA commitments. It also frees operational teams to focus on exceptions rather than routine tasks.

Application and wider value

A unified, cross-CSP assurance layer makes new commercial models viable. With this solution, CSPs can offer SLA-guaranteed interconnect and cross-border services because they can trace and resolve issues across partner networks with shared logic. Enterprise can buy experience tiers with confidence that response, diagnosis, and remediation work consistently even when multiple CSPs are involved. This creates space for premium offerings and new revenue sources. As summarized by KDDI's project lead Fukumoto Norihiro, "the Catalyst has resulted in 50% faster mean time to repair (MTTR), 90% reduction in alarm noise, and 60% OPEX reduction in NOC operations."

The OpEx impact is also significant. The B2B2X model, delivered through TM Forum APIs on the ODA platform, reduces the need for each provider to build and maintain its own bespoke assurance stack. Shared standards cut integration effort, while automated OAM flows remove the manual coordination that normally slows cross-provider operations. As autonomous agents shoulder more of the diagnostic burden, CSPs can reallocate skilled teams to higher-value engineering and innovation tasks.

More broadly, the Catalyst demonstrates how AI can address one of the hardest problems in telecom operations: fault detection and recovery across fragmented network domains. The challenge intensifies when incidents span multiple CSPs, yet this is where the industry needs the most reliability. By showing that agentic AI, shared context, and open standards can close this gap, PRISM AI gives operators a practical blueprint for future assurance systems and for collaboration at scale.

The achievements of the project, supported by rigorous research including work by Japan’s NICT, point towards a future where cross-domain resilience is not an exception but a baseline expectation. As the industry evolves toward autonomous networks, solutions like PRISM AI demonstrate that coordination, clarity, and intent-driven action can transform how CSPs maintain service quality, protect customer experience, and grow new business with confidence.