With strategy defined and architecture aligned, the real challenge for communication service providers (CSPs) is execution at scale. TM Forum addresses this by providing industry-validated tools, reference implementations, skills programs and collaborative projects that move operators from planning to production.
Accelerating AI innovation through collaboration
TM Forum’s Innovation Hub is a collaborative R&D space uniting global telecom companies, tech giants, and solution providers. The goal is to reduce risk while accelerating delivery of production-ready solutions. Projects capture agile, entrepreneurial thinking.
The Forum’s Catalyst projects are fast-paced, proof-of-concept initiatives in which CSPs, vendors and hyperscalers co-create solutions, typically moving significantly faster than with in-house R&D. “Moonshot” Catalysts are the top tier, tackling the industry’s most ambitious challenges, including GenAI, autonomous networks and sustainability.
Embedding AI in industry-validated frameworks
As agentic AI becomes embedded across networks and IT operations, TM Forum advocates for operationalizing AI by leveraging Open Digital Architecture (ODA), moving AI from fragmented experimentation to scalable, open production systems. The emphasis is on making AI “native” to the architecture, using standardized, governed data and APIs to avoid vendor lock-in.
“We must not rush to implement a new technology, otherwise we risk creating data silos and data debt, as well as technical debt,” cautioned Guy Lupo, EVP of AI and Data at TM Forum, in a blog post discussing AI-native transformation. He defined data debt as what happens when “proprietary AI models, infrastructure dependencies and geopolitical constraints lock critical insights and data sets into immovable systems.”
Autonomous networks: a key priority
Autonomous networks are a major focus for CSPs. TM Forum’s six-level maturity model (Level 0 to Level 5) guides operators from manual operations to full autonomy, delivering demonstrable advancement through faster service rollouts, improved network performance, lower operational complexity and more predictable customer experience.
China Mobile: Achieving Level 4 autonomy at scale
China Mobile, operator of the world’s largest 5G network with 578 million subscribers, reached Autonomous Network Level 3.2 by 2023. Advancing to Level 4 required a shift to true autonomous decision-making by eliminating human intervention in operations and maintenance wherever possible, reducing site visits to one per ticket and resolving faults before they affect customers.
Working with Huawei, China Mobile backs TM Forum’s “Make AN Level 4 Real” Moonshot project, which targets a 30% reduction in operating expenditure, or OpEx. Their solution, which won the TM Forum Excellence Award for Autonomous Networks, uses a large AI model with more than 10 billion parameters, autonomous closed-loop operation tool chains, and two types of AI agents: role-based copilots and scenario-based agents that handle fault management and complaint handling through natural language.
To scale nationally, China Mobile mapped its full range of operation and maintenance scenarios, then piloted in select provinces before replicating city-by-city and province-by-province. Success was measured using TM Forum’s key business value indicators (KBI), key effectiveness indicators (KEI) and key capability indicators (KCI).
Results from the initial two-province rollout included improved fault management, reduced mean time to repair, or MTTR, and a reduction in operational manpower equivalent to 5,500 people per year: 75% in network operations and 25% in customer operations.
“With L4 network autonomy as the goal, Huawei and China Mobile are co-creating the future,” said Sam Wang, General Manager of ADN Solution at Huawei. “We are jointly innovating more Level 4 best practices and systematically accelerating China Mobile’s AN journey.”
Telstra’s ‘knowledge plane’ innovation
Australian operator Telstra has pioneered a new architectural approach to network service automation, centered on a concept called the “knowledge plane.” Rather than writing automation scripts for every conceivable scenario, an approach that doesn’t scale, Telstra encodes engineers’ domain knowledge into knowledge graphs up front.
“What we’ve done is fundamentally different,” said Telstra Chief Architect Mark Sanders. “The knowledge around what the network domain is and how the resources function is encoded in knowledge graphs right up front. That’s the big shift.”
The knowledge plane spans multiple network domains, enabling outcome-based, intent-driven operations rather than step-by-step processes. To build it, Telstra created a telco-specific ontology using TM Forum’s Information Framework, a domain-specific language, or DSL, for interacting with the knowledge graphs, and a graph-aware application powering a streamlined orchestrator.
A key component is the Network Differencing Engine, which compares current service states against desired outcomes and enables precise, intent-based actions, effectively making the network self-healing. As principal engineer Matt Beanland explains: “The model-driven differencing engine is the navigator, providing turn-by-turn instructions based on your goals and policy. The orchestrator is the driver, executing those instructions.”
Telstra’s innovative approach is the result of a broader transformation effort. The company achieved “running on ODA” status in 2023 and continues to leverage TM Forum’s Open Digital Architecture (ODA) and Open APIs to develop its Telstra Reference Architecture Model (TRAM).
Telstra also drew on TM Forum’s Guiding Principles for Building and Measuring Autonomous Network Solutions — which shaped its approach to observability and responsibility, delineating machine versus human roles — and digital twin assets, since the Knowledge Plane mirrors the live network as a virtual representation of services, configurations and dependencies.
The business impact has been significant: Service creation and onboarding time dropped from 12-18 months to three months, order-processing time fell by 30%, and Telstra’s Net Promoter Score improved. The company is now preparing to launch a capability that lets customers design their own network services through an intuitive interface, something unthinkable with legacy systems.
“Our goal is not just to make this for Telstra,” Sanders says, “but to cast the net wider for the industry.”