Services
Network Solutions and Services
Reimagining 5G and network solutions for future businesses
Highlights
- Autonomous networks progress through stages from assisted automation to fully self-operating systems, but most operators today are in intermediate stages, where automation exists but remains siloed.
- Most wireless operators have automated parts of network operations within individual domains such as RAN and core, where systems can detect issues and trigger predefined corrective actions automatically.
- However, the absence of real-time cross-domain correlation in networks is a critical barrier to realizing matured autonomy, [AA3.1][NG3.2]what we call autonomous network (AN) maturity.
- A cross-domain correlation fabric spanning unified telemetry ingestion, semantic normalisation, a multi-modal correlation engine, and intent-driven actuation is an architectural prerequisite for self-managing, intent-driven networks.
- Agentic AI, combining domain-level agents with an orchestrator that holds network intent, converts cross-domain correlation from a monitoring capability into real-time autonomous closed-loop action, a defining capability of AN maturity.
- Global system integrators with deep telecom domain expertise can play a key role in delivering cross-domain correlation as a vendor-neutral, outcome-accountable capability.
On this page
On this pageinpage
The autonomy goal
Most operators have automated aspects of network operations within individual domains such as RAN and core. The next threshold demands coordination across these domains.
Wireless operators worldwide have invested significantly in autonomous network (AN) programmes, guided by the TMForum Autonomous Networks framework. Domain-level progress is real: Self-organising radio access network (RAN) functions can optimise antenna configuration autonomously; core fault management correlates alarms and triggers scripted remediation; and provisioning workflows run with minimal human touch. Most operators today operate at AN level 2 or level 3 with closed-loop automation [AA4.1][NG4.2]established within individual domains, but siloed between them.
Closed-loop automation is the ability to monitor network conditions, detect issues, determine the likely cause, and trigger corrective action automatically within a defined scope.
At the next stage of maturity, the network must act across domains as a coordinated system, understanding end-to-end conditions and responding against a defined business intent with minimal human intervention.
The jump from autonomy level 3 to level 4 [AA5.1][NG5.2]is qualitatively different. It requires the network to perceive conditions spanning domain boundaries, reason about cross-domain causality, and execute a coordinated response aligned with a stated intent without human facilitation at each step. Cross-domain correlation, the ability to sense, reason, and act across domains without waiting for manual coordination at every step, [AA6.1][NG6.2]is the foundational capability that makes this possible. A network that cannot see across its own domain boundaries cannot act across them.
Where operatrors are
RAN-Core correlation is progressing though transport, business support system (BSS), and fixed domains remain largely siloed.
At the RAN-Core interface, there have been several advancements in terms of correlation. The Network Data Analytics Function (NWDAF), introduced through standards from the 3rd Generation Partnership Project (3GPP), allows analytics across the radio and core network. The Open-Radio access network (O-RAN) non real time radio intelligent controller (Non-RT RIC) and non-radio application (rApp[AA7.1][NG7.2]) architecture [AA8.1]in 5G networks further enables cross-layer intelligence. These are production-deployed capabilities at a growing number of operators running standalone 5G architectures.
The correlation perimeter, however, remains narrow. These are promising developments, but they still cover only part of the problem and do not provide true end-to-end operational visibility.
Customer experience signals from a business support system (BSS) typically flow on batch cycles, not real-time feeds. For converged operators, fixed and mobile domains remain unconnected: The broader problem is that transport, customer systems, and fixed networks are still often managed in silos. A transport bottleneck may not be visible until radio performance degrades. A fixed network issue may appear as a mobile demand spike, causing the network to react to the wrong problem.
The operational consequence is misattributed causality and avoidable service degradation.
The correlation gap
The business case for closing the correlation gap is clear: it reduces mean time to repair (MTTR), protects SLAs, and enables outcome-based commercial models.
Major incidents in wireless networks are disproportionately multi-domain in nature. The coordination overhead of bridge calls, manual cross-OSS correlation, and inter-team root cause negotiation represents the majority of elapsed MTTR in complex incidents. Autonomous cross-domain correlation eliminates this overhead. Root cause is established algorithmically in seconds, with structured evidence ready for immediate execution. Network operations center (NOC) shift from manual pattern matching to exception handling and continuous improvement.
As enterprise 5G deployments proliferate with private networks, network slicing for industrial verticals, guaranteed-latency services, end-to-end SLA commitments span RAN, transport, and core simultaneously. Cross-domain correlation is the only mechanism that can assure these commitments in real time. Beyond quality, cross-domain visibility is the evidential foundation for outcome-based commercial models: proving performance against contractual obligations and identifying where shortfalls originate. In a market shifting from effort-based to outcome-based pricing, that commercial capability has direct revenue implications.
Correlation fabric
We envisage the correlation fabric architecture as four interdependent layers—each a prerequisite for the next.
The first layer is unified telemetry ingestion — collecting real-time data from all network domains into a shared streaming environment.
The second layer is semantic normalisation — ensuring data from different systems uses a common structure, meaning, and time interval.
The third layer is the correlation engine — identifying patterns, likely causes, and operational risk using rules and AI.
The fourth layer is actuation — executing approved network actions in line with defined intent and policy.
Agentic layer
Domain agents and an orchestrator agent convert correlated insight into autonomous cross-domain action.
An AI agent in network management perceives its environment through telemetry, reasons using defined objectives, plans a response, and executes autonomously within defined boundaries. The correlation fabric produces insight; the agentic layer converts that insight into action. Domain agents operate at the network domain level: the RAN agent understands self-organizing network (SON) parameters and load-balancing levers; the transport agent understands routing policies and capacity thresholds across fronthaul, midhaul, and backhaul; the core agent understands UPF placement and slice policy; the BSS agent understands SLA commitments and regulatory obligations. Each domain agent manages within-domain closed loops and exposes a queryable interface enabling live domain state to be interrogated in real time.
Above the domain agents, an orchestrator agent holds network intent. When the correlation engine detects a cross-domain condition, for instance, rising transport congestion threatening an enterprise slice SLA, the orchestrator queries domain agents for available levers, reasons about the optimal combination of actions, arbitrates conflicts, sequences execution, and monitors outcomes. This model allows domain-level intelligence to be coordinated at the network level, so actions remain aligned to overall service objectives rather than optimising one domain in isolation. The agentic maturity arc is phased. In the near term, agents deliver cross-domain root cause narratives to analysts in seconds. In the mid-term, they propose action plans for human authorisation. At AN level 4, they execute autonomously within pre-cleared intent boundaries. At this stage, the network, in defined and growing scope, runs itself. This path to autonomous cross-domain execution is outlined in the section below.
Maturity roadmap
We visualize a three-phase journey from correlation foundation to autonomous cross-domain execution, with each phase gated by measurable outcomes.
Phase 1 — Foundation: This involves deploying streaming telemetry across all domains into a unified data lake; establishing CMDB accuracy as an active programme workstream; implementing the semantic normalisation layer; and deploying rule-based cross-domain correlation for priority failure scenarios.
Success criterion: Cross-domain root cause identification validated against historical incidents.
Phase 2 — Agentic insight: In this phase, the focus is on extending domain agents to transport and BSS; deploying orchestrator agent with cross-domain synthesis capability; and enabling agentic recommendation with human authorisation for execution.
Success criterion: Recommendations accepted for more than 70% of qualifying incidents with measurable MTTR reduction.
Phase 3 — Autonomous execution: At this stage, operators need to strengthen the guardrail framework; grant orchestrator agent cross-domain actuation authority for pre-cleared actions; implement the intent repository with machine-readable service-level objectives; and extend autonomous scope progressively by domain based on confidence and adherence metrics. Success criterion: More than 80% of qualifying cross-domain incidents resolved autonomously within guardrail boundaries with full audit trail compliance.
Organizational model
Technical architecture alone will not deliver cross-domain autonomy—the operating model must transform alongside it.
Traditional NOC structures—separate RAN, Core, transport, and BSS operations teams with distinct SLA accountability and change management authority—are architecturally incompatible with cross-domain closed loops. AN level 4 requires either a unified operations function with cross-domain execution authority, or a federated model in which automated arbitration supersedes domain-level decision making for defined event types. Intent ownership—the senior role responsible for the commercial and operational objectives the autonomous system pursues—does not exist in most operator organisations today and must be deliberately created.
Trust in autonomous action is built through shadow-mode operation: The system proposes actions that humans execute, enabling quality measurement before authority is progressively transferred. Cross-domain autonomous actions must also satisfy change management frameworks and regulatory obligations—automated change records, immutable audit trails, and pre-cleared action taxonomies are prerequisites. Actions affecting emergency services infrastructure or lawful intercept capabilities require specific pre-clearance and escalation design before autonomous authority is granted.
SI's role
While equipment vendors play a critical role within their respective domains, enabling cross-domain correlation as a cohesive and integrated capability requires a broader, multi-domain perspective.
A system integrator plays a key role. It operates without domain allegiance, selecting the best-fit analytics platform, normalisation approach, and orchestration layer independent of equipment vendor relationships. The system integrator also brings deep multi-vendor integration engineering like normalising proprietary telemetry schemas across multiple vendors, linking topology across management systems never designed to interoperate, and maintaining CMDB accuracy across vendor boundaries. In other words, it does all the hard engineering that makes the correlation fabric work at production scale.
The system integrator provides programme leadership across four parallel workstreams—technical architecture, data governance, organisational change, and commercial framework—that must advance in concert for each phase gate to be reached. In managed operations, the system integrator is the natural custodian of the agentic layer, maintaining agent quality, extending guardrail scope, and operating the continuous learning loop that improves correlation models from resolved incidents.
In conclusion, cross domain correlation is the foundational capability required to achieve fully autonomous networks. Operators have already made meaningful progress within individual domains, but genuine autonomy will depend on breaking these silos and acting across them as one coordinated system.
The path forward is clear. It is phased, measurable, and achievable. Operators that build cross-domain visibility, align it with intent, and progressively enable autonomous execution will improve operations today while creating the foundation for long-term competitive advantage.