Login
Register
Member Insights
With autonomous networks poised to unlock hundreds of millions in opex savings for telcos, Shamik Mishra, CTO of Connectivity at Capgemini Engineering, looks at why a trusted data fabric is a prerequisite.
Shamik Mishra, Capgemini
AI agents need a real-time data fabric to enable autonomy at scale
Autonomous networks could unlock $150 million to $300 million in opex savings per operator over the next five years, and most large telecom operators have set their sights on Level 4 autonomy by 2030. However, nearly half remain stuck between Level 2 and 3. Agentic AI has emerged as a crucial enabler for closing that gap, but operators may be overlooking another critical element: trusted intelligence.
As operators turn to AI agents for advanced monitoring, execution and orchestration, they must recognize that agents rely on data and intelligence to understand context, weigh decisions and operate within system guardrails. For many telecom organizations, the path to higher network maturity therefore isn’t just a matter of building and orchestrating agents, but creating a robust underlying data fabric that powers decisions, coordinates actions and ensures agents are acting with the right context.
Traditional telecom data architectures, including data lakes and warehouses, were designed to look backward. While they are effective for managing structured, predictable datasets, they cannot support the real-time, closed-loop decision-making required by agentic AI.
This gap becomes critical as operators pursue higher levels of network autonomy. Unlike legacy networks where control depends on predefined rules, architectures that are autonomous and enabled by AI agents depend on context. Agents infer decisions based on the accuracy and consistency of the data they receive. If that context is stale, fragmented or inconsistent across domains, actions may not align with network needs and goals.
Unfortunately, many communications service providers (CSPs) do not yet have this data foundation. In a TM Forum survey of 87 senior IT and network executives, only two said their organizations had achieved fully democratized, cross-organizational data integration, underscoring an almost universal need among operators for data modernization.
Modern data fabrics have five core technical requirements:
A real-time data fabric provides the contextual layer agents need to act on trusted network intelligence across radio access network (RAN), core, transport and operational support system (OSS) environments. But to support autonomy at scale, that fabric cannot sit outside the architecture as a supporting capability. As recommended within TM Forum’s Open Digital Architecture (ODA), data must be elevated to a first-class component with clear ownership across production, transformation, exposure and governance.
This means ODA components must remain loosely coupled but tightly aligned through shared data models and standardized APIs. Without that alignment, agents operating across different domains risk acting on inconsistent context or triggering conflicting decisions.
This model keeps data quality close to the teams that understand each network domain best, while enforcing “FAIR” principles that make data findable, accessible, interoperable and reusable. It also strengthens digital twin capabilities by ensuring that simulations and decision intelligence run on accurate, synchronized data.
BT Group offers a real-world example. Its data mesh architecture organizes data streams from multiple sources, while a single data fabric applies access controls across the enterprise. Combined with composable IT components built on ODA, this AI-ready foundation has helped BT manage variable data streams and use network topology and infrastructure data to accelerate fiber and 5G rollouts.
For telecom organizations, the operational priority is to prepare the real-time data foundation that makes model autonomy viable. Here we outline five steps to reach Level 4 autonomy:
1: Establish domain data ownership
To break down organizational silos, operators should assign data ownership to domain-specific teams and make them responsible for publishing their data as standardized, curated products.
2: Transition to streaming telemetry
Operators must replace legacy, batch-based pipelines with real-time streaming collectors that instantly recognize network state changes and prevent closed-loop breakdowns.
3: Align with standardized data models
To enable cross-domain reasoning and ensure consistent action, operators must normalize incoming data streams against standardized information models, such as the Information Framework and Intent Specifications (TR299).
4: Build contextual digital twins
Operators should introduce continuous state models to synchronize real-time network events, establish persistent state awareness, and turn isolated telemetry into actionable insights.
5: Treat data governance as operational risk management
Operators should establish robust governance frameworks and leverage standardized guardrails, such as those defined in the TM Forum AI-Native Blueprint, to ensure that agent interactions occur within a secure, verifiable framework.
Autonomous networks can unlock significant efficiency and growth for telecom operators, but Level 4 autonomy will not be achieved through agents alone. For CSPs, the next phase of network transformation depends on building the data foundation that enables autonomous action.
With a robust AI-native data fabric in place, operators can move beyond isolated pilots and fragmented automation toward production-grade autonomy, accelerating the network transformation journey and building the maturity that drives business impact.
