The telecommunications sector has officially breached the boundaries of theoretical automation and entered the era of operational autonomy. [1] For decades, managing complex telecom networks meant dealing with siloed data, delayed troubleshooting cycles, and manual infrastructure configurations that took weeks to deploy. Today, a paradigm shift is occurring as global technology leaders move from rigid automation to autonomous agentic architectures.
Two massive milestones highlight this evolution. First, Nokia and Microsoft have successfully automated complex network configurations using a combined Data Suite and Fabric framework managed by AI agents. Concurrently, on live wireless networks, ZTE and XLSMART have successfully tested Level 4 autonomous network optimization agents, providing the industry’s first concrete validation of high-level closed-loop automation.
Together, these breakthroughs solve two of the largest bottlenecks in modern telecommunications: the agonizingly slow process of multi-vendor data integration and the operational inefficiency of manual radio frequency optimization.
1. The Nokia-Microsoft Framework: Architecting the Agentic Data Foundation
The persistent hurdle in deploying AI across telecom networks has never been a lack of algorithms—it has been a lack of clean, unified, and timely data. Telecom infrastructure generates massive lakes of telemetry, but this data is scattered across multi-vendor silos and disparate software domains. Conventional extraction, transformation, and loading (ETL) pipelines can take weeks to prepare a dataset for consumption. By the time the data is ready, the network event it describes has long passed.
The Nokia and Microsoft collaboration shatters this bottleneck by establishing an agentic, unified data foundation. The architecture fuses Nokia Data Suite’s ready-to-use telco data products with the cloud-scale analytics, governance, and reasoning capabilities of Microsoft Fabric.
+-------------------------------------------------------------+
| NOKIA DATA SUITE |
| (Ready-to-use Data Products + TM Forum Semantic Models) |
+------------------------------+------------------------------+
|
v (Staged in Minutes, Not Weeks)
+-------------------------------------------------------------+
| MICROSOFT FABRIC |
| (OneLake Storage + Unified Governance Analytics) |
+------------------------------+------------------------------+
|
v (Drives Downstream Reasoning)
+-------------------------------------------------------------+
| AGENTIC AI SOLUTIONS |
| [Predictive Maintenance] [VoNR Assurance] [Fault Mgmt] |
+-------------------------------------------------------------+
The Technical Architecture: Data Products to OneLake
Rather than forcing telecom providers to manually construct data pipelines, the framework links Nokia’s modular data products directly into Microsoft Fabric’s OneLake storage environment.
- Semantic Modeling: Nokia packages raw network feeds using standardized, TM Forum-aligned telecom semantic modeling. This provides the underlying data with “glass-box” structural quality verification.
- AI Readability: Because the data is contextualized with telecom semantics before ingestion, downstream AI agents can interpret the relationships between different network entities automatically.
- Speed of Insight: This tight coupling reduces data preparation timelines from weeks down to mere minutes. AI applications no longer analyze historical artifacts; they feast upon query-ready, real-time operational data.
Empowering Multi-Agent Closed-Loop Operations
With an agentic data foundation established, software agents can run continuous, autonomous operations across live, multi-vendor footprints. They can correlate network infrastructure anomalies with enterprise IT records or third-party data on demand.
Initial production applications for this framework include:
- Autonomous Voice-over-New-Radio (VoNR) Assurance: AI agents monitor the quality of 5G voice services in real time, executing instant optimizations if call drop rates threaten specific thresholds.
- Geo-Experience Analysis: Tracking user-perceived performance metrics across specific geographic coordinates to identify micro-bottlenecks.
- Predictive Maintenance and Automated Root-Cause Analysis: Detecting subtle structural anomalies and diagnosing hardware vulnerabilities prior to equipment failure, prompting immediate self-healing scripts.
Furthermore, because the solution deploys smoothly across cloud, hybrid, and on-premises environments, regional operating units can easily customize their deployments to align with stringent data sovereignty regulations.
2. ZTE and XLSMART: Delivering Verifiable Level 4 Autonomy
While Nokia and Microsoft focused on building the data fuel for autonomous engines, ZTE and XLSMART targeted the physical execution edge. At the ZTE Global Summit & User Congress, the two entities showcased a milestone achievement: the validation of a Network Optimization Agent deployed in live, commercial wireless operations.
This deployment represents a definitive shift toward Level 4 (L4) Autonomous Networks, as defined by the TM Forum and the wider telecom industry.
Unpacking Level 4 Autonomy
The evolution of autonomous networks is categorized into levels, mirroring the automotive industry’s approach to self-driving vehicles. While Level 3 networks feature conditional automation where a human must step in to handle anomalies, Level 4 autonomy introduces fully closed-loop execution within specific domains.
In an L4 ecosystem, the AI agent is given absolute operational control over defined scenarios—such as wireless coverage optimization. The system handles perception, deep analysis, decision-making, and command execution completely on its own. Human intervention is minimized, allowing the network to continuously morph and repair itself dynamically based on live environmental fluctuations.
The Live-Network Test Metrics
The joint live-network trial deployed ZTE’s agent software directly into XLSMART’s commercial infrastructure to automate wireless coverage tuning. Rather than depending on manual drive-testing or delayed engineer analysis, the Network Optimization Agent monitored live cells, calculated target coverage alterations, and applied changes dynamically.
The trial yielded a series of highly measurable business and performance indicators:
- 25% Closed-Loop Closure Rate: Within the validated scope, the AI agent successfully resolved 25% of all coverage optimization events entirely without human oversight, from initial detection to final validation.
- Slashing Manual Workloads: By offloading a quarter of core network tuning events to software agents, manual engineering hours dropped drastically, accelerating the optimization cycle from days to seconds.
- Cross-Domain Orchestration: Combining single-domain autonomous agents with intelligent cross-domain orchestration significantly lowered the Mean Time to Repair (MTTR) and systematically reduced the need for physical, on-site technician visits.
3. Direct Comparison: Two Approaches to Network Intelligence
The achievements of these two partnerships represent complementary forces driving the telecommunications sector toward a unified vision of zero-touch operations.
| Strategic Dimension | Nokia & Microsoft Collaboration | ZTE & XLSMART Partnership |
|---|---|---|
| Primary Objective | Building a unified, agentic data foundation across the telecom stack. | Live-network validation of high-level closed-loop automation. |
| Core Technologies | Nokia Data Suite + Microsoft Fabric (OneLake). | ZTE Network Optimization Agent + Cross-Domain Orchestrators. |
| Operational Scope | Cross-domain analytics, data readiness, and multi-vendor integrations. | Targeted single-domain execution transitioning to cross-domain fault location. |
| Key Achievement | Reduced data preparation times from weeks to minutes for machine reasoning. | Attained a 25% autonomous loop closure rate in live commercial wireless operations. |
| Impact on Engineers | Minimizes manual data manipulation, providing immediate, trusted insight pipelines. | Minimizes manual intervention, dropping MTTR and eliminating routine on-site visits. |
4. Technical Enablers: Fueling the Autonomous Transition
The dual success of these projects underscores that network autonomy is no longer restricted by software limitations. It relies heavily on three core technological pillars:
- Telco-Specific Semantic Fabrics: General-purpose AI models cannot run telecom infrastructure without translation layers. Both frameworks succeed because they apply strict industry standards (like TM Forum semantics) directly to the data layer, ensuring that incoming telemetry maps perfectly to real-world network dependencies.
- Explainable Machine Reasoning: For operators to give AI systems administrative write-access, the agent’s logic must be transparent. Nokia’s “glass-box” data verification ensures that every operational input fed to Microsoft Fabric tools is clean, structured, and auditable, removing the risk of unpredictable “black-box” decisions.
- Advanced Closed-Loop Perception: ZTE’s L4 architecture succeeds due to its continuous telemetry monitoring capabilities. The agent doesn’t check network parameters periodically; it perpetually perceives environmental variations, processes the optimal response via internal reasoning engines, and applies patches instantly.
5. Challenges on the Path to Full Level 5 Dark NOCs
While these milestones are monumental, achieving the holy grail of a fully autonomous, human-free “Dark NOC” (Network Operations Center) requires overcoming substantial hurdles.
Multi-Vendor Integration Stalls
Nokia and Microsoft have proved that multi-vendor data synthesis is possible, but real-world telecom topologies are incredibly fractured. Legacy hardware from various eras often utilizes proprietary protocols that do not export data cleanly. Scaling these agentic frameworks globally requires absolute multi-vendor cooperation, which remains a friction point in a highly competitive hardware market.
The Cross-Domain Fault Bottleneck
An anomaly in user experience might appear as a RAN layer degradation, but its root cause could be a physical routing failure deep in the backhaul or a cloud configuration mismatch in the core data center. While single-domain agents (like ZTE’s coverage optimization agent) excel within their specific boundaries, cross-domain fault location remains incredibly complex. If independent agents across multiple layers misinterpret each other’s parameters, they run the risk of creating conflicting configuration changes that can destabilize system states.
Regulatory and Sovereignty Concerns
Telecom networks are classified as critical national infrastructure. Handing complete control to autonomous software networks forces carriers to satisfy intense regulatory scrutiny. Frameworks must be meticulously designed to isolate data regionally, ensuring that international AI models do not violate domestic privacy or localized data sovereignty policies.
Conclusion: The Programmable, AI-Native Future
The successful automation of intricate network configurations by Nokia and Microsoft, paired with ZTE and XLSMART’s live validation of Level 4 autonomy, shifts the conversation from if networks can run themselves to how fast operators can scale these deployments.
We are moving definitively away from static, rigid infrastructures that require constant manual intervention. The industry is rapidly building programmable, AI-native platforms capable of perceiving, thinking, and acting on their own. As these systems scale, the telecom network of tomorrow will operate like a living digital ecosystem—constantly analyzing its own health, adjusting its resources within minutes, and self-healing at machine speed.
Autonomous Network Milestones (FAQ):
Q1: What did Nokia and Microsoft achieve in their collaboration?
- A: Nokia and Microsoft successfully unified and automated telecom data workflows by linking Nokia Data Suite with Microsoft Fabric. This integration reduces the time required to extract, format, and prepare multi-vendor network telemetry for AI consumption from weeks to just a few minutes, creating a ready-to-use foundation for operational AI agents.
Q2: What is the significance of the trial conducted by ZTE and XLSMART?
- A: ZTE and XLSMART successfully validated an autonomous Network Optimization Agent on a live, commercial wireless network. The agent managed real-world radio coverage tuning entirely on its own, proving that high-level autonomous network management can step out of the lab and function reliably in live consumer environments.
Q3: What defines a “Level 4” autonomous network?
- A: According to the industry standard taxonomy, Level 4 (L4) autonomy means the network features fully closed-loop execution within specific domains. The AI agent can perceive anomalies, analyze options, make decisions, and execute configuration changes completely without human oversight in defined scenarios, minimizing manual intervention.
Q4: How did the ZTE and XLSMART agent perform in live testing?
- A: The live-network test achieved a 25% closed-loop closure rate. This means the autonomous agent successfully handled and resolved a quarter of all radio coverage optimization issues from start to finish without any human intervention or engineering support.
Q5: Why are TM Forum semantic models critical to the Nokia-Microsoft framework?
- A: General AI models cannot natively understand raw router logs or radio telemetry. Nokia uses TM Forum-aligned semantic modeling to label and structure data with standard telecom terminology. This transforms raw logs into a “glass-box” readable layout that Microsoft Fabric-hosted AI agents can immediately comprehend and reason against.
Q6: What are the main challenges to scaling these solutions across entire carrier networks?
- A: The primary challenges include:
- Legacy Multi-Vendor Friction: Integrating older infrastructure that relies on closed, proprietary data protocols.
- Cross-Domain Complexities: Coordinating single-domain agents (like RAN or Core) so their automatic adjustments do not accidentally cause conflicting, disruptive configurations across other network layers.
- Data Sovereignty: Meeting strict national regulatory policies regarding critical infrastructure control and user privacy.
Q7: Will Level 4 autonomy eliminate the need for human Network Operations Center (NOC) engineers?
- A: No. Instead of replacing engineers, it upgrades their responsibilities. Engineers shift away from manual, repetitive log parsing and manual radio tuning. They step into higher-level roles as Policy Architects, defining the safety envelopes, compliance rules, and performance budgets within which the autonomous agents must operate.
#AutonomousNetworks, #TelcoAI, #NokiaDataSuite, #MicrosoftFabric, #ZTEGlobalSummit, #XLSMART, #NetworkAutomation, #ZeroTouchOperations, #TelecomInnovation