Unlocking the next frontier of Local AI on Windows for Telecommunications
At Build 2026, we announced significant new capabilities across our on-device AI platform that expand what’s possible at the edge, in the field, and in the network operations center. This post covers two key areas and what they mean for telecom: new capabilities in Microsoft Foundry on Windows, and the introduction of secure agent execution and governance on Windows. Before we go further, let’s be clear about one thing: this is not a cloud-versus-device story. Cloud remains essential for training, orchestration, and shared intelligence across the network. What on-device AI does is expand the operating model. It adds a new execution tier that unlocks scenarios that were harder to deliver before because of latency, resiliency, privacy, or runtime constraints. The goal is to place AI where it creates the most value, and for many telecom workloads, that’s closer to the point of action. What Local AI unlocks for telecom When you bring AI inference to the device, it opens up scenarios that may have been limited in cloud-only architectures: Customer care at the point of interaction. Agents that can respond faster, with richer context, without round-tripping to a data center. Field and frontline continuity. Workflows that keep working when connectivity is weak, inconsistent, or unavailable: remote cell sites, underground facilities, rural deployments. Time-sensitive network operations. Decisions that need to happen quickly at the edge without cloud latency. Privacy and sovereignty by default. Inference on sensitive telemetry and customer data without it leaving the jurisdiction. Local agents that act, not just advise. A shift from generic copilots to agents that take action on the device, with better support for long-running execution, tool use, and reliable runtime behavior. Windows is where much of telecom’s enterprise and operational work already happens. NOC workstations, field technician laptops, back-office systems, retail point-of-sale. That installed base creates an opportunity: make the existing device a better surface for local AI execution without requiring new hardware or new deployment footprints. The differentiator isn’t just model runtime. It’s the complete environment: model execution, agent runtime, containment, and governance, integrated at the OS level. At Build 2026, we shipped the next wave of platform capabilities that make this concrete. Microsoft Foundry on Windows: new capabilities at Build 2026 Microsoft Foundry on Windows is our unified local AI platform, first introduced at Build 2025 and generally available since earlier this year. It gives developers and operations teams three complementary ways to run AI on-device, with hardware abstraction across NPU, GPU, and CPU. At Build 2026, we announced new capabilities across each layer of the stack that significantly expand what’s possible for telecom workloads. Windows AI APIs: expanding reach and capability Windows AI APIs provide the fastest path to integrating local AI into applications. These are ready-to-use, OS-managed capabilities powered by on-device models. These are the same capabilities that power Copilot+ local AI experiences today. Outlook Summarize, Alt Text Generation, and Teams video upscaling all run through these APIs locally on Copilot+ PCs. When you build with Windows AI APIs, you’re tapping into the same proven inference pipeline that millions of Windows users already rely on daily. New at Build 2026: We are expanding Windows AI APIs beyond NPUs to CPUs and GPUs, bringing local AI experiences to a much broader set of Windows 11 devices. We’re also introducing a new Speech Recognition API (aka.ms/speech-recognition-api) for real-time on-device speech-to-text. Unlocked potential for telecom: Field technicians use on-device OCR to read equipment serial numbers, fiber labels, and antenna configurations offline at remote cell sites. No connectivity, no cloud dependency, no delay. NOC analysts run local semantic search across alarm logs and trouble tickets for faster root-cause analysis, without sending sensitive network data off-premises. The new Speech Recognition API enables on-device transcription of field calls, dispatch audio, and customer interactions, even without network connectivity. This unlocks new workflows for dictation-enabled maintenance reports and real-time captioning in operations centers. With GPU and CPU expansion, these capabilities now reach the broader fleet of Windows devices already deployed across telecom operations, beyond Copilot+ PCs. Learn more at aka.ms/WinAI/APIs. Foundry Local: flexible local inference on standard enterprise hardware Foundry Local is Microsoft’s cross-platform local AI runtime that lets developers run AI directly into their applications without cloud dependency and cloud token costs. First previewed at Build 2025 and generally available since April 2026, Foundry Local is designed for scenarios where you need more flexibility than a pre-built API: running open-source language and vision models on-device with full control over inference and acceleration. Foundry Local supports five SDKs (C#, Python, JavaScript, Rust, and C++) and works on standard enterprise hardware. It is not limited to Copilot+ PCs. It handles model packaging, runtime fragmentation, hardware differences, and deployment complexity so teams can focus on shipping features rather than stitching infrastructure together. With access to the Foundry model catalog, developers can run models for chat completions, embeddings, live audio transcription, and vision-language tasks entirely on-device. New at Build 2026: The 1.2.0 release expands language support in the Live Transcription API, introduces inference cancellation for cleanly stopping in-flight requests, and upgrades to Windows ML (WinML) 2.0 for NPU and GPU acceleration without extra installation steps. Cross-region catalog routing now delivers faster model downloads for end users. Foundry Local also adds a Responses API for structured agentic interactions including tool calling and multimodal vision-language input. Foundry Local is already being used across privacy-sensitive, performance-sensitive, and hardware-diverse scenarios. From local assistants and document workflows to multimodal context collection and enterprise AI pipelines, developers are using it to reduce platform complexity and deliver production-ready AI experiences faster. Learn more at devblogs.microsoft.com/foundry. Unlocked potential for telecom: Data sovereignty. Run inference on network telemetry, customer records, and performance data without it leaving your data center or jurisdiction. Critical for regulated markets and government contracts. Autonomous network operations. Deploy local models that triage alarms, correlate events across network domains (transport, IP, radio), and recommend remediation actions, all on-prem with enterprise-grade privacy and offline operation. Disconnected scenarios. Field crews in rural tower deployments or underground facilities get full AI-assisted workflows without depending on backhaul connectivity. Custom model flexibility. Fine-tune models on your proprietary network data (topology, KPIs, fault patterns) and deploy them locally through Foundry without vendor lock-in. Windows ML: the foundation for local AI at scale Windows ML, generally available since September 2025, is the native ONNX Runtime-based engine for deploying custom AI models across Windows devices. It acts as a hardware abstraction layer to easily distribute models across CPUs, GPUs, and NPUs while ensuring the best on-device performance. Windows ML is also the foundational layer that powers Foundry Local, providing broad silicon support and a consistent deployment experience. At Build 2026, Windows ML continues to expand its silicon support and serves as the engine delivering unmetered intelligence on Windows, enabling the new Aion models and expanded API reach described above. Unlocked potential for telecom: Predictive maintenance. Deploy custom ONNX models trained on your specific equipment failure data to field devices or edge nodes. Predict faults before they become outages. RF optimization. Run propagation models or beamforming optimization algorithms on-device at cell sites for real-time RAN tuning without cloud latency. Anomaly detection. Deploy bespoke network anomaly detection models, trained on your traffic baselines, to edge servers for instant threat and fraud identification. Bring-your-own-model. CSPs with mature data science teams deploy proprietary models across their Windows fleet without being constrained by a single vendor ecosystem. Local AI Telecommunications App demo We’ve built a sample field technician application that combines Windows AI APIs (on-device OCR for reading equipment labels) with Foundry Local (a workflow that diagnoses the fault, recommends a fix, and generates a work order), all running locally with zero cloud dependency. (Note: This is NOT a Microsoft licensed product or application. This is NOT an officially supported Microsoft offering. Microsoft assumes no responsibility or liability for any use of this code.) GitHub: https://github.com/opesam/LocalAIDevelopment/tree/telecom-field-ai Introducing Aion: a new family of models for local AI At Build 2026, we introduced Aion, a new family of on-device models purpose-built for local execution on Windows. Aion represents a new generation of small language models that are smaller, faster, and more efficient than our previous Windows OS SLMs. Over time, Aion models across multiple parts of the on-device AI stack, though they have not yet been fully productized into specific APIs or features. Try Aion at aka.ms/tryaion The Aion family includes two models at launch: Aion 1.0 Instruct is optimized for instruction-following tasks: summarization, text generation, classification, and general-purpose language understanding. It is smaller and more efficient than our previous Windows OS SLM, delivering improved quality at lower compute cost. Aion 1.0 Plan is a 14-billion parameter reasoning and tool-calling model with 32K context, designed for agentic workflows. It can reason over user intent, invoke tools, manage files, and orchestrate sub-agents, enabling fully agentic execution on-device without cloud dependency. These models point toward a future where the full local AI stack benefits from purpose-built, efficient on-device intelligence. As they are integrated across APIs and features, they will unlock new classes of experiences for developers building on Windows Unlocked potential for telecom: A path to agentic network operations. As Aion models mature and integrate across the platform, they open the door to local agents that can triage alarms, correlate events across network domains, and recommend remediation actions entirely on-prem. On-device planning and orchestration. Aion’s reasoning and tool-calling direction enables multi-step workflows: a field technician’s device diagnosing a fault, looking up parts inventory, generating a work order, and scheduling a follow-up, all locally without waiting for a cloud round-trip. Efficient local intelligence. Aion’s smaller footprint and improved efficiency mean these capabilities will run well on the existing Windows hardware already deployed across telecom operations, without requiring top-tier GPU resources. Secure agent execution and governance on Windows Running models on-device is one thing. Running agents on-device, agents that take actions, access systems, and operate continuously, is a different challenge entirely. The question every CSP asks is the same: how do we let agents operate with enough autonomy to be useful, while maintaining the governance and security posture our business demands? At Build 2026, we introduced new capabilities that address this directly. Introducing Microsoft Execution Containers (MXC) Microsoft Execution Containers (MXC), introduced at Build 2026, is a policy-driven execution layer integrated into Windows and WSL. Developers declare what an agent can access (files, network, system resources) and MXC enforces those boundaries at runtime. It is not a single isolation technology. It is a composable spectrum: from lightweight process isolation to session isolation to full micro-VMs, dynamically matched to the risk profile of the workload. Get started at github.com/microsoft/mxc. Every agent running inside MXC is bound to an identity, either local or cloud-provisioned through Entra, so every action is attributable, auditable, and governed at the kernel level. MXC is delivered as an SDK, now available in early preview. OpenAI and NVIDIA are already building on it. Unlocked potential for telecom: OSS/BSS protection. Constrain agents that interact with production network management systems to read-only access. Prevent accidental or malicious configuration changes to live network elements. Multi-vendor agent isolation. When you deploy AI agents from multiple vendors (RAN analytics, transport optimization, independent fault correlation), each operates within strict boundaries. No cross-contamination of data or access between vendor agents. Regulatory auditability. Every agent action is attributable to a governed identity. Essential for telecom regulatory compliance: data retention, lawful intercept boundaries, GDPR, local privacy regulations. Zero-trust agent architecture. MXC enables a zero-trust posture for AI agents: assume breach, verify continuously, limit blast radius. This maps directly to the security frameworks telecom already operates within (3GPP, NIST CSF). Edge deployment security. Agents running at cell sites or multi-access edge compute nodes operate within kernel-enforced boundaries, even in physically less secure environments. Learn more at: Windows Platform Security for AI Agents and aka.ms/BUILD_SecurityBlog Windows 365 for Agents: enterprise governance for agent fleets Windows 365 for Agents, generally available since May 2026, provides computer-using agents with secure, managed Cloud PCs to execute enterprise workflows. It is the enterprise-grade platform for discovering, governing, and securing AI agents at scale, whether those agents are built by Microsoft, third parties, or your own teams. New at Build 2026: Windows 365 for Agents now integrates natively with MXC on Windows, delivering Defender, Entra, Intune, and Purview protections so that security and IT teams can constrain and secure local agents to prevent enterprise risk. This integration is available in preview starting July 2026. Windows 365 for Agents provides end-to-end observability across your entire agent estate. It surfaces “shadow AI”: agents that have been deployed across the organization without IT visibility or approval. Agents run in secure, managed Cloud PC environments (learn more), fully separate from user machines, for executing enterprise workflows that require a full desktop context. Unlocked potential for telecom: Agent fleet management at scale. CSPs may operate hundreds or thousands of AI agents across network domains: core, transport, access, customer operations. Windows 365 for Agents provides a single pane of glass for visibility and governance. Think of it as a network management system, but for AI agents. Shadow AI discovery. Identify agents that business units or vendors have deployed without IT approval, a growing risk in large telecom organizations where multiple vendor ecosystems operate simultaneously. Policy-driven autonomy. Define precisely what each class of agent can do. A network healing agent can restart services but cannot modify routing tables. A customer service agent can access CRM but cannot process billing adjustments above a threshold. Compliance and data governance. Purview integration ensures that agents handling customer PII (call detail records, location data, subscriber information) comply with GDPR, local telecom regulations, and data residency requirements. Operational continuity. If an autonomous agent misbehaves in a live network, operators can detect, halt, and remediate before customer impact. Mission-critical for the five-nines availability telecom demands. Why Local AI matters now for Telecommunications Local AI is not a retreat from cloud. It is an expansion of what AI can do across the telecom operating model. Build 2026 delivered the next wave of capabilities CSPs need to move from experimentation to execution: expanded local inference that reaches more devices and more scenarios, a new generation of on-device models purpose-built for agentic reasoning, and for the first time, OS-level containment and governance for autonomous agents. Microsoft Foundry on Windows gives you the inference engine: pre-built APIs for quick wins, an open-source runtime for agentic workflows, and a custom model deployment layer for your proprietary workloads. MXC and Windows 365 for Agents give you the guardrails: containment, identity, policy, and observability for every agent in your fleet. Telecom has always operated at the intersection of scale, reliability, and trust. The scenarios where on-device AI creates the most value (customer operations, field workflows, network operations, autonomous remediation) are exactly the scenarios where latency, resiliency, privacy, and governance matter most. The question for CSPs is no longer whether to run AI locally. It’s which workloads benefit most from this new execution tier, and how quickly you can get started. Get started Visit the Microsoft booth at DTW Ignite 2026 in Copenhagen (June 23–25) Reach out for a Local AI pitch and envisioning conversation: aka.ms/opesamolowu Explore Microsoft Foundry on Windows: learn.microsoft.com/windows/ai Get started with MXC: github.com/microsoft/mxc Learn more about Windows 365 for Agents: Windows 365 for Agents documentation Read the full Build 2026 Windows platform blog: Build 2026: Furthering Windows as the trusted platform for developmentThe Case for an Ontology Layer in Telecoms
The Challenge: Data Without Context in the Telco AI Era GenAI and analytics tools and solutions pull data from applications and from downstream and external data sources. However, when data is decontextualized from the applications where it was created, it often loses the business meaning, relationships, and rules that were embedded in the original application. This information is critical for producing dependable, business-relevant output using AI. For example, in the sales domain, customers might be called an account, a consumer, in the network domain will be mostly called subscriber. Different business units might use a different denomination. Line of business systems might have their own naming conventions; external entities might follow telecom-specific standards. Telecom operators are increasingly constrained not by lack of data, but by lack of data meaning: Network data (RAN, core, transport). IT/OSS/BSS data. Customer interaction data (apps, call centers, digital channels). External ecosystem data (partners, devices, edge). While these datasets are abundant, they are: Fragmented across systems. Largely unstructured, with proprietary formats or inconsistently modeled. Needs improvement in terms of governance in terms of quality and semantics. As a result: Humans find challenging to decipher the meaning, or to extract insights - for example: what are the numbers of consumers being affected by a drop of the QoS KPI p95_latency_ms that affects apps latency in real time. AI models struggle to interpret relationships across domains. Use cases (e.g., autonomous networks, customer 360, fraud detection) require heavy manual data engineering. Scaling GenAI and agent-based workflows becomes complex. This reflects a broader issue: when data is extracted from its originating system, it loses its business meaning, relationships, and operational rules. This semantic consistency creates a shared language across the enterprise and ecosystem, independent of specific formats or proprietary structures. The ontology manages translation, meaning, and relationships, keeping business logic simple and portable. As a result, it provides the foundation for AI to reason, plan, and execute consistently. The ontology (preview) item (part of the Fabric IQ (preview) workload) digitally represents the enterprise vocabulary and semantic layer that unifies meaning across domains and OneLake sources[1]. “Ontology modeling recognizes business concepts over table schemas[2].” Why Telcos need an Ontology Some organizations are investing in a robust ontological layer to address this problem. A semantic layer is a system of technologies and techniques that organizes and maintains a consistent and unified representation of data from various sources that is interpretable by humans and machines. Functioning conceptually between data and the user (human or machine), a semantic layer describes data’s business meaning, its relationships with other data, and the rules the data must adhere to. “Microsoft Fabric IQ provides a way to define business vocabulary in an ontology and bind the ontology to data sources[3].” An ontology layer in telecom functions as a “digital understanding layer” for enterprise data: Describes network entities (tower cells, sites, devices, slices). Defines customer concepts (subscriber, household, product bundles). Encodes relationships (e.g., customer → service → network resource). Helps to enforce business and regulatory rules (e.g., data privacy, SLA constraints). This layer enables: Cross-domain reasoning (network ↔ customer ↔ service). More consistent AI outputs. Faster onboarding of new data sources and use cases. Figure 1 - An ontology (preview) related to mobile networks Critically, in the GenAI era, AI requires machine-readable context (e.g., ontologies, knowledge graphs) to generate reliable outputs at scale. “In Microsoft Fabric IQ, ontology modeling differs from traditional analytical modeling by starting with business concepts rather than specific use cases[4].” Your ontology is a graph database—a data structure built from two fundamental elements: nodes and edges. In the graph view below, each colored circle is a node (a real record—a base station, a network, and each labeled arrow connecting them is an edge (a relationship). Figure 2 - An ontology (preview) represented as a graph that can be queried by a data agent The target state for an ontology layer for Telecom AI GenAI raises the bar for accessing contextual metadata at speed and scale. Without that context, GenAI tools are less able to produce relevant, reliable responses for business users and are more likely to misinterpret data. To address this, organizations need modern capabilities such as semantic search, which goes beyond keyword matching by using metadata to understand the intent behind a query. They also benefit from knowledge graphs, which provide a machine-readable representation of enterprise knowledge for applications such as network anomaly agents, recommendation engines, fraud detection, and customer experience agents. An ontology layer captures an organization’s understanding of data, regardless of where it resides or how it is structured. In the GenAI era, it must hold the context, including data usage constraints, permitted combinations, personal data handling rules, and compliance requirements. It can also encode expertise such as data quality judgments and approved exceptions to business rules. With this semantic layer, organizations can scale AI without redesigning governance for each use case, improving the accuracy of AI agents and GenAI models while reducing risk. Microsoft Fabric IQ offers the following capabilities[5]: Ontology items define your business vocabulary and bind concepts to data sources. Data agents answer natural language questions across multiple data sources. Graph stores and queries connected data using relationships. Semantic models provide a starting point for generating ontologies from existing Power BI models. Figure 3 - A data agent querying data that uses a Microsoft Fabric IQ ontology as its data source A modern telecom ontology should: Represent enterprise knowledge. Describe the elements that form the network topology. Service portfolio hierarchies. Customer relationships. Unified metadata across OSS, BSS, and IT domains. Enable AI-Driven Data Access Semantic search over telecom data (intent-based vs keyword-based). Context-aware for GenAI (prb_util_pct above 80, means RAN congestion and a troubleshoot agent needs to be enacted). Agent interoperability via shared ontology. Encoding Governance and Compliance Data usage policies (e.g., GDPR, NIS2, lawful intercept constraints). Permitted data combinations. Data access rules for AI agents. Capture Telco Domain Expertise Codified engineering rules (e.g., network thresholds, fault correlations). Data quality scoring and trust levels. A well-designed ontology layer transforms telecom data from fragmented technical signals into actionable, AI-ready knowledge assets. Microsoft’s strategic partners, such as Kenmei, Infosys and Tech Mahindra, are already innovating and bringing in the market ontology-driven solutions for the telecom industry. Conclusion For telecom operators, investing in a robust semantic layer means building data assets that can be understood and trusted at speed and scale. In the GenAI era, this machine-readable representation of enterprise knowledge is important for enabling AI to interpret data, reason effectively, and create operational advantage. Ontology offers major transformative potential for telecoms, with the ability to reshape operations in ways few other technologies can. While it may seem abstract and at times misaligned with initiatives such as data estate modernization, it is a foundational capability for operators to move toward AI-native operations. The semantic consistency, contextual intelligence, and action-oriented structure provides that it is difficult to be replicated by traditional data management approaches. [1] https://learn.microsoft.com/en-us/fabric/iq/ontology/overview [2] https://learn.microsoft.com/en-us/fabric/iq/ontology/overview [3] https://learn.microsoft.com/en-us/training/modules/understand-fabric-iq-fundamentals/ [4] https://learn.microsoft.com/en-us/training/modules/understand-fabric-iq-fundamentals/ [5] https://learn.microsoft.com/en-us/training/modules/understand-fabric-iq-fundamentals/Reimagining Network Operations: How Microsoft NetAI Tackles Hyperscale Challenges
The Business Imperative: Why Network Operations Must Change Modern network operators face a perfect storm of challenges: Exponential Growth in Events and Maintenance Network events and maintenance activities are increasing at an unprecedented rate. According to Microsoft’s analysis, weekly network events are projected to grow from hundreds to thousands over the next five years. Maintenance activities are expected to follow a similar trajectory. Without automation, this growth would require a dramatic—and unsustainable—increase in staffing. Rising Operational Costs Dense Wavelength Division Multiplexing (DWDM) operations, which are critical for high-capacity fiber networks, are both costly and labor-intensive. The global spend on Network Operations Center (NOC) services exceeds $5 billion annually, with total network operations costs reaching $250 billion per year. As networks expand, these costs threaten to spiral out of control. Human-Centered Workflow Limitations Manual processes are slow, error-prone, and unable to keep pace with the scale and speed of modern networks. Organizational inertia, fragmented tooling, and siloed systems further impede efficiency. Engineers are often bogged down by device-specific command-line interfaces and isolated management systems, slowing onboarding and cross-functional collaboration. Safety and Reliability Concerns Early attempts to automate network operations with AI revealed critical gaps. Traditional AI models struggled with limited context, leading to unpredictable outcomes and eroding trust. Machine learning models often generated false positives, overwhelming operations teams with unnecessary alerts. The risk of unsafe command execution—where an autonomous agent might inadvertently disrupt service—remained a constant concern. The Talent Crunch As network complexity grows, so does the demand for skilled engineers. Yet, hiring and training enough talent to keep up with operational demands is neither cost-effective nor sustainable. The industry faces a widening gap between operational needs and available expertise. NetAI: A Strategic Shift Toward Autonomous Operations Microsoft NetAI is not just another automation tool—it’s a strategic framework for transforming how networks are managed. By leveraging intelligent agents, curated context, and modular workflows, NetAI enables the Azure Networking team to move from reactive, manual processes to proactive, AI-driven automation. Key Objectives of NetAI Achieve Fully Autonomous Network Operations: NetAI aims to eliminate the need for manual intervention in the majority of network incidents, allowing intelligent agents to detect, diagnose, and resolve issues independently. Minimize Human Involvement in Incident Lifecycle: By automating detection, root cause analysis, and repair, engineers can focus on higher-order tasks like agent enablement and system design. Scale Operations Without Scaling Headcount: As network events grow exponentially, NetAI maintains a flat staffing curve by automating repetitive and time-consuming tasks. Ensure Deterministic and Reliable AI Behavior: NetAI emphasizes deterministic workflows, engineered prompts, and stateful context management to guarantee consistent and safe outcomes. Enable Role-Based Agent Collaboration: Specialized agents operate within defined scopes, enhancing reliability and accountability. Support Organizational Transformation: NetAI redefines the role of network engineers, shifting their focus from manual operations to automation enablement and system governance. How NetAI Addresses Business Challenges NetAI’s architecture is designed to tackle the most pressing operational challenges head-on: Scalability: By automating incident handling, NetAI enables organizations to manage more events without increasing headcount. Cost Efficiency: Automation reduces the need for expensive, labor-intensive operations, delivering significant cost savings. Reliability and Safety: Deterministic workflows, strict guardrails, and role-based access controls ensure that automation is both reliable and safe. Organizational Agility: By freeing engineers from repetitive tasks, NetAI empowers them to focus on innovation and strategic initiatives. A summary table from the whitepaper highlights the breadth of challenges addressed, from exponential event growth and DWDM inefficiency to fragmented tooling and repair delays. The Measurable Benefits of NetAI The impact of NetAI on Microsoft’s global network operations has been transformative. Here are some of the most notable outcomes: 40% More Incidents Handled Per Person AI agents manage detection, diagnosis, and resolution, allowing engineers to handle 40% more incidents per person. This shift enables human operators to focus on higher-value activities such as agent enablement, prompt refinement, and system design. 80% Faster Root Cause Analysis With agents like Pal leveraging topology, telemetry, and historical data, the time required to isolate and understand complex issues has dropped by 80%. This acceleration not only improves service reliability but also reduces the operational burden on Tier 2 support teams. 25% Reduction in Time to Repair (TTR) Autonomous agents like Miles initiate and manage fiber repair workflows without waiting for human coordination, streamlining the resolution process and minimizing service disruption. Flat Staffing Curve Despite 10x Event Growth Perhaps most impressively, NetAI has enabled Microsoft to maintain a flat staffing curve even as the number of incidents and maintenance tasks has increased dramatically. This decoupling of scale and headcount is a critical advantage in hyperscale environments. Improved Consistency and Reliability Deterministic automation reduces false positives and operational noise, improving consistency and reliability across the board. Cultural Transformation Beyond the numbers, NetAI has fostered a cultural shift within Microsoft. Engineers are no longer just responders—they are automation architects, designing and refining the systems that drive autonomous operations. This evolution enhances job satisfaction, reduces burnout, and positions the workforce for long-term success in an AI-driven future. Strategic Collaboration and Industry Impact The success of NetAI is not just a product of internal innovation—it’s also shaped by Microsoft’s active collaboration with network operators around the world. Through joint workshops, pilot deployments, and feedback loops, Microsoft works closely with partners to tailor the agentic framework, workflow orchestration, and safety protocols to real-world conditions. This collaborative approach accelerates the maturity of NetAI while empowering operators to modernize their network operations. To further accelerate adoption, Microsoft has introduced the Network Operations Agent (NOA) Framework—a deployment and enablement toolkit that packages NetAI’s best practices, engineered prompt libraries, architectural blueprints, and modular components into a reusable format for operators. As NetAI continues to evolve, Microsoft is focused on expanding agent roles, enhancing multi-agent coordination, and deepening integration with operational systems. The vision is clear: smarter, safer, and more strategic operations that redefine what’s possible in network management. Download the Full Whitepaper Ready to dive deeper? The full Microsoft NetAI whitepaper explores the strategic vision, technical architecture, and real-world impact of autonomous networking. Download it here to learn how your organization can benefit from the next generation of network operations: ⬇️Download the Microsoft NetAI WhitepaperEvolving the Network Operations Agent Framework: Driving the Next Wave of Autonomous Networks
The original announcement of the Network Operations Agent (NOA) Framework outlined a bold mission: provide telecom operators with a modular, multiagent foundation to accelerate the journey toward autonomous networks. NOA combined intelligent agents, unified data access, and strong governance to help operators modernize complex, cloud scale environments. A follow-up is timely. Over the past year, customer deployments, industry collaboration, and Microsoft’s own internal learnings have fueled a rapid evolution of the framework. Telecom operators face skyrocketing event volumes, rising operational costs, and a deepening skills gap—conditions echoed in the challenges documented by Microsoft’s own NetAI. Against this backdrop, NOA’s enhancements are designed to deliver more automation, more intelligence, and more resilience—while maintaining safety and human oversight. What’s New: Key Enhancements in NOA v2 Adoption of NetAI Concepts and Best Practices Although NOA is not NetAI, its latest iteration incorporates proven architectural principles and operational practices from the NetAI program—including intelligent agents, curated context, engineered prompts, and deterministic workflows. NetAI’s focus on scaling automation without scaling headcount, and its multi‑agent roles such as diagnostics and fiber repair, informed several NOA improvements. This infusion strengthens NOA’s ability to support autonomous incident detection, diagnosis, and remediation while preserving telco-grade safety and predictability. And the partnership is symbiotic, as NetAI is leveraging NOA’s modern, Foundry based architecture in future iterations. The UI for AI: Deep Integration with Microsoft 365 Copilot and Teams NOA now embraces Microsoft Teams, Outlook, and Copilot as the primary user interface for AI agents—turning everyday collaboration tools into a real-time operations cockpit. Operations engineers can ask questions such as “What’s causing the latency spike in region X?” directly in Teams and receive agent generated diagnostics, summaries, or recommended actions. Agents proactively post alerts, help draft incident summaries, and allow supervisor approvals—merging human and AI workflows seamlessly. This “UI for AI” approach dramatically reduces friction, shortens response cycles, and boosts adoption across operations teams. Migration to Microsoft Foundry & the Microsoft Agent Framework A cornerstone update is NOA’s full alignment with Microsoft Foundry technology, including: Microsoft Agent Framework – An extensible, open‑source orchestration layer providing standardized agent communication, tool use (via MCP), A2A, observability, and hybrid deployment options. Foundry Agent Service – A secure runtime for deploying and scaling NOA’s multi‑agent workflows. Foundry Observability – Built-in memory, traceability, and monitoring for every action and interaction. Foundry IQ – Intelligence services enabling context retrieval, semantic grounding, and safe decision-making. These enhancements give NOA a more deterministic, governed, and auditable operational backbone—crucial for regulated telco environments. Integration of TM Forum Open APIs via MCP The NOA Framework’s alignment with TM Forum standards has expanded significantly since the initial release. In addition to its broader support for TM Forum’s Autonomous Network principles, NOA now incorporates TM Forum Open APIs for trouble ticketing, delivered through the Model Context Protocol (MCP) integration layer. A key enhancement is the explicit support for the TMF621 Trouble Ticket Management API, the industry standard interface for creating, updating, querying, and resolving trouble tickets in OSS/BSS environments. By exposing TMF621 operations through NOA agents and MCP toolchains, the framework enables: Automated creation of standardized TMF621‑compliant trouble tickets Agent driven correlation of telemetry, diagnostics, and ticket history Seamless interoperability with existing OSS/BSS and service desk systems Consistent, vendor neutral workflows for incident lifecycle management This deeper TMF621 alignment ensures NOA agents can participate directly in ticket-driven operational processes while maintaining full compliance with telco grade open standards. Strengthened Security, Governance, and Compliance Controls NOA’s governance model has been expanded with: Read-only defaults and restricted permissions, enforced through managed identities and RBAC Detailed action logging for auditability Operator defined policy gates requiring human approval for sensitive tasks Support for hybrid and multi-cloud deployments with consistent identity and compliance This ensures telcos can deploy autonomous agents without compromising regulatory responsibilities or operational safety. Real-World Impact: How Azure Accelerates Autonomous Operations NOA’s evolution is driven not only by design, but by field usage across operators and large‑scale networks. Microsoft’s internal Azure Networking success Microsoft used NetAI to automate fiber incident management across the global Azure backbone—achieving: 60% reduction in time-to-detect fiber issues 25% improvement in repair times These measurable gains demonstrate the potential of agentic operations at hyperscale. These learnings, best practices, concepts, and designs are at the core of NOA. Alignment with broader telco transformations Far EasTone Telecom (FET) exemplifies how leading operators are turning this architecture into real operational impact. FET is leveraging the NOA framework to redefine cloud native network operations by embedding agentic AI across its NOC and change management workflows. Today, nearly 60% of its NOC operations are AI-assisted, with about 10,500 operational tasks executed per month, including incident summaries, automated ticket closure, network checks, and proactive voice notifications. AI agents now handle largescale alarm correlation and root cause analysis in seconds, supporting nearly 7,000 monthly operational queries with an average response time of 16 seconds, and enabling most maintenance actions to complete within one minute. This shift has significantly reduced human error, accelerated recovery times, and allowed engineers to focus on higher value work. Built on Azure cloud native and hybrid data principles, FET can scale network intelligence securely across on premises and cloud environments, deploying capabilities closer to customers while maintaining carrier grade reliability, performance, and regulatory confidence—demonstrating how adaptive cloud and AI can turn network operations into a strategic advantage. Vodafone is working with Microsoft to apply a proven AI‑powered blueprint for autonomous network operations across transport infrastructure and field‑force management. The collaboration combines Vodafone’s deep network expertise with Microsoft Foundry and the NOA framework to modernize how large‑scale telecom networks are operated. This blueprint is built on Microsoft’s own experience running autonomous agents across its global Azure transport network, where AI continuously monitors performance, identifies root causes, and autonomously manages more than 65% of fibre‑break field dispatches—improving time to repair by up to 25% and accelerating root‑cause analysis by 80%. By applying these proven capabilities to Vodafone’s transport network, the two companies are accelerating the shift toward intelligent, automated transport network operations across the telecom industry. “By working with Microsoft, we’re combining deep network expertise with proven AI‑powered operations to create something greater than either could achieve alone. Together, we’re building intelligent, automated transport network operations that empower our teams and deliver faster, more resilient connectivity networks for our customers.” —Alberto Ripepi, Chief Network Officer, Vodafone Other operators, including AT&T, T-Mobile, Telefónica, and MEO, are adopting Microsoft Foundry as a blueprint for scaling agentic AI across complex, multi-vendor networks. Case studies from global operators leveraging Foundry and Azure AI capabilities—though not NOA specific—demonstrate similar patterns of AI driven operational gains: AT&T unifies and analyzes massive volumes of network data with Microsoft Azure—particularly Azure Databricks, Power BI, and cloud-scale AI analytics—in a secure lakehouse architecture, enabling faster AI-driven insights, improved network-informed decision-making, and more efficient, scalable operations across its telecom network. T‑Mobile leverages Microsoft Azure, data, and AI analytics platform—combining services like Azure Data Explorer, Azure Databricks, and AI-driven analytics—to ingest and analyze trillions of network data points in near real time, giving teams deep visibility into network performance, proactively identifying and resolving issues, and continuously optimizing the customer experience across its nationwide network. Telefónica España applies Microsoft Azure’s big data, AI, and automation capabilities—such as Azure Data Explorer and Azure Databricks—to analyze massive volumes of network data in real time, detect anomalies, and enable self‑optimizing 4G/5G networks that improve performance, reliability, and customer experience while reducing operational costs. MEO uses Microsoft AI systems to improve technician efficiency and transparency in operational processes. These customer achievements reinforce the architectural direction taken with the NOA Framework, leveraging Azure data and AI capabilities like Azure Databricks, Fabric, and Foundry to operate world-class networks. Architecture & Key Components Based on the latest Foundry capabilities, the NOA reference architecture emphasizes four key subsystems, detailed in the next sections. UI for AI The UI for AI is the human interaction layer through which operators engage with the NOA system using natural language. In the NOA architecture, this UI is delivered through familiar enterprise surfaces—the WebApp, Microsoft Teams, and Microsoft 365 Copilot—allowing users to initiate workflows, review agent findings, approve actions, and observe outcomes without switching tools or learning new interfaces. Teams based agent interactions and supervisor controls now form a key architectural pillar. What it enables Conversational interaction with the NOC Manager and specialist agents Human-in-the-loop approvals for diagnostics, remediation, and ticket actions Consistent experience across web, Teams, and Copilot while preserving enterprise identity and permissions Agentic Governance Agentic Governance is the policy, safety, and control layer that enforces Responsible AI, security, compliance, and observability across all NOA agents and workflows. In the NOA architecture, this governance is provided by the Foundry Control Plane and associated guardrails, evaluations, audit logs, and operator views. What it enables Runtime guardrails for content safety, PII protection, prompt injection, and tool misuse Human-in-the-loop escalation and approval controls Auditability, compliance reporting, and policy enforcement across all deployed agents Foundry Control Plane Foundry Control Plane functions as the governance + observability + safety enforcement layer that sits above the multi-agent system, ensuring agent workflows can run in production with the right controls. It is where governance, observability, security, and Responsible AI controls are enforced for agent workflows. It’s presented as the mechanism that turns the solution into a scalable, governed, and enterprise-ready operations framework. Concretely, the Foundry Control Plane provides: Guardrails (safety controls): a central place to apply and manage protections against harmful content, PII leakage, prompt injection, and off-topic drift—plus the ability to create custom guardrails tailored to telecom needs (e.g., restricting sensitive config exposure or limiting high-risk tool calls during incidents). Evaluations (quality and readiness checks): benchmarking agents via “evaluation runs” using built-in evaluators (accuracy, safety, coherence, domain quality), helping validate integrations (like MCP / Fabric) and catch regressions before rollout. Operator view for fleet monitoring: centralized monitoring across deployed agents/channels, including fleet health (uptime, errors, sessions), performance (latency, token usage, throughput), and compliance signals (guardrail violations, policy alerts). Asset inventory and lifecycle management: a unified inventory of deployed agents, models, and tools (e.g., MCP servers, Fabric Data Agents), supporting versioning, staging/rollback, usage analytics, credential rotation, and model policy enforcement. Compliance management: centralized management of guardrails, policy templates, audit logs, and risk dashboards to produce audit-ready governance across the agent estate. Foundry Control Plane is the control-and-monitoring system that makes autonomous, multi-agent incident workflows safe, auditable, and operable at enterprise scale. Agent 365 Agent 365 functions as the enterprise management plane for AI agents—the place where agents are registered, controlled, observed, and secured at scale. Agent 365 is positioned as the central control plane that provides: Unified registry / inventory of AI agents (a single place to manage what agents exist across the enterprise). Access control so only the right users/roles can use particular agents and tools. Observability to monitor agent usage and behavior across deployments. Enterprise security controls applied consistently across agents and where they run. In short, Agent 365 is the “management hub” for governing and operating the organization’s agent estate, complementing (and working alongside) the Foundry Control Plane’s runtime guardrails/evaluations and the broader multi-agent orchestration. Agentic Framework The Agentic Framework is the orchestration layer that enables stateful, multi‑agent workflows within the NOA system. Built on the Microsoft Agent Framework, it allows a NOC Manager (Niobe) to coordinate specialized agents (e.g., Troubleshooting, Telemetry, Ticketing, Field Ops) through delegated tasks, shared context, and ordered execution. What it enables Vertical orchestration of specialized agents under a single supervisory control Durable workflows spanning diagnosis, remediation, validation, and ticketing Agent-to-Agent (A2A) communication and integration with external agents (e.g., ServiceNow) Microsoft Agent Framework The Microsoft Agent Framework functions as the foundation layer for building the Telco NOA solution’s stateful, multi-agent system, providing the structure for how agents are defined, orchestrated, and operated end-to-end. Specifically, the Microsoft Agent Framework enables: Stateful, multi-agent workflows for NOC operations: it underpins the NOA platform, letting agents collaborate across a single incident while retaining shared context over time (not just one-off prompts). A centrally orchestrated (“vertically orchestrated”) model: a NOC Manager (Niobe) manages task delegation, sequencing, and shared context across the workflow, while specialized agents (Troubleshooting, Incident Management, Field Ops, SONiC, Security & Compliance) execute domain-specific tasks. Secure cross-platform agent collaboration via A2A (agent-to-agent) communication: external agents (example given: ServiceNow Now Assist) can plug into the ecosystem through A2A to enable coordinated actions across platforms. Operationalization of multi-agent systems with durable workflows: an open-source SDK used for designing, orchestrating, and operationalizing multiagent systems with durable, stateful workflows. All of that positions Microsoft Agent Framework as the core multi-agent runtime/SDK layer that makes the NOA incident flow (supervisor + specialist agents + external A2A agents) possible and maintainable as a workflow-driven system. Foundry Workflows Foundry Workflows function as the orchestration layer that defines, executes, and governs the end-to-end sequence of multiagent actions required to resolve network incidents in the Telco NOA Framework. The workflow models how tasks flow across agents—from initial intent capture through diagnostics, remediation, ticketing, verification, and closure—under the supervision of a central orchestrator agent. Specifically, the workflow encodes the ordered handoffs and decision logic between agents such as the Supervisor Agent (Niobe), Network Telemetry Analyzer, Troubleshooting Agent (Pal Locke), SONiC Agent, Field Ops Agent, and Ticketing Agent. Rather than relying on static scripts or point integrations, the workflow graph visually and operationally represents how the Supervisor dynamically delegates tasks, routes context, and coordinates execution across specialized agents. Foundry Workflows also provide a testable and observable execution environment. Workflows can be previewed in a sandbox mode, allowing presenters to simulate real incident flows, trigger agent interactions via natural language prompts, and validate that tool calls (for example, MCP-based ServiceNow or TM Forum APIs) execute correctly before production deployment. During execution, the workflow enables full traceability and auditability. Debug and conversation detail views expose each step in the workflow, including routing decisions by the Supervisor, tool invocations, intermediate responses, and final outputs. This makes workflows not just an automation mechanism, but a governance artifact that supports troubleshooting, cost analysis, and compliance review. Finally, Foundry Workflows act as the deployment unit for operationalizing agentic solutions. Once validated, the same workflow can be published with one click to Microsoft Teams or Microsoft 365 Copilot, preserving guardrails, evaluations, RBAC, and monitoring. This allows the exact same orchestrated logic to run consistently across chat, Copilot, and custom channels without re‑engineering. Telco IQ Telco IQ is the intelligence layer that grounds agent reasoning in telecom specific operational knowledge. In the NOA architecture, this intelligence is delivered through Foundry IQ and Fabric IQ, which provide retrieval augmented reasoning across structured telemetry, operational data, and unstructured domain knowledge. What it enables Telecom aware reasoning using KPIs, SOPs, policies, and historical incidents Multi-hop retrieval and citation backed answers Reduced hallucinations through grounded, policy aware context Fabric IQ Fabric IQ functions as the semantic intelligence layer that helps agents make better decisions by organizing operational + analytical data into business concepts, so the agents can reason over it more effectively (not just retrieve raw records). This dramatically simplifies connecting to data while improving the quality of the results. Foundry IQ Foundry IQ functions as the unified knowledge and retrieval layer that grounds the NOC agents with relevant, policy-aware context during incident troubleshooting. Foundry IQ is described as: A Foundry Knowledge Base (built on Azure AI Search) that unifies content such as network security policies, troubleshooting guides, and SOP knowledge articles across sources including Microsoft 365/SharePoint, Fabric IQ/OneLake, Azure Blob Storage, ADLS Gen2, ServiceNow tables, and the web. The capability that automates key RAG pipeline steps—including query planning, multi-hop reasoning, and answer synthesis with citations—so each agent doesn’t have to rebuild chunking, indexing, and connector logic per project. A mechanism for enterprise-grade security and governance, explicitly leveraging Entra ID governance and respecting Microsoft Purview sensitivity labels, while reducing hallucinations through grounded retrieval. The knowledge backbone agents rely on in the workflow: for example, the NOC Agent leverages the Foundry IQ knowledge base to retrieve operational insights and summarize likely causes, and other specialized agents (e.g., troubleshooting and security compliance). Universal Data Access Universal Data Access is the data foundation that unifies real-time, structured, and unstructured data sources into a single, governable knowledge fabric for NOA agents. The architecture explicitly combines Microsoft Fabric Eventhouse, Azure Cosmos DB, ADLS Gen2, Azure Blob Storage, ServiceNow tables, and Microsoft 365 sources, all accessed through governed tools and identity passthrough. What it enables Real‑time telemetry access for diagnostics and validation Persistent conversational memory and incident history Secure, identity aware access across enterprise and operational systems NOTE: While the NOA Framework sample implementation and demo utilize this specific data architecture, the Framework can integrate into any existing data environment. The focus of NOA is to simplify and accelerate the development of network AI agents by leveraging your existing data estate. With the included Microsoft Fabric connectors and OneLake virtualization, NOA can reason over: Real‑time telemetry OSS/BSS data Ticketing systems Multi-cloud or on-prem data stores These updates collectively make NOA more scalable, more open, and easier to operationalize across diverse network environments. Meet the Agents The following agents are the core “cast” of the NOA. Each one represents a specialized capability (telemetry analysis, troubleshooting, device interaction, compliance validation, ticketing, and field operations) coordinated by a central supervisor. NOC Manager Agent (Niobe) The central orchestrator and human interface: manages task delegation, sequencing, and shared context across workflows; coordinates the other specialized agents through Foundry’s runtime. This agent flags/triages critical tickets, coordinates with connected agents, and uses Foundry IQ to retrieve operational insights and summarize likely causes for the operator. Network Telemetry Analyzer Agent Analyzes network performance metrics (e.g., packet loss, throughput) and uses the Fabric Data Agent to generate/execute KQL queries against the Fabric (Eventhouse/KQL DB) to support diagnostics and post-mitigation verification. Network Troubleshooting Agent (Pal Locke) Runs diagnostic playbooks and remediation steps; retrieves insights from Foundry IQ and proposes fixes via the SONiC Agent for device-level commands. SONiC Agent Interfaces directly with the network elements (e.g. SW-TOR-05 device) to execute commands and retrieve system-level data/telemetry (e.g., OS details, PFC watchdog stats, logs). Network Security Compliance Agent Validates post-mitigation QoS/SLA metrics (e.g. latency and jitter) using Foundry IQ, ensuring fixes remain within defined SLAs. Ticketing Agent Manages incident tickets and documentation: creates/updates tickets, logs agent actions, supports auditing/handovers, and integrates via an MCP server exposing ServiceNow tools and TM Forum Trouble Ticket Open API to retrieve history for accurate categorization and updates. Field Ops Agent (Miles Dyson) Handles physical/field issues (e.g. fiber cuts) and collaborates with the appropriate teams via email. Building the Autonomous Network Ecosystem NOA is intentionally designed as a partner-extensible framework, not a closed product. As a result, many partners are adopting NOA as a reference implementation for agentic operations—then integrating it into their existing agent framework, OSS/BSS, assurance, and automation portfolios to deliver differentiated autonomous network solutions for operators. How partners are adopting NOA Starting point for agentic architecture: using NOA’s supervisor + specialist agent pattern as the baseline for incident, problem, and change workflows. Accelerator for telco-grade governance: adopting the guardrails, approvals, observability, and auditability concepts to meet operator safety and compliance expectations. Reference integration blueprint: mapping NOA’s tool and API-driven approach (MCP, TM Forum Open APIs) onto their own integration adapters and connectors. Blueprint for the “UI for AI” in operations: embedding agents into Teams/Copilot experiences to reduce swivel-chair work and drive practitioner adoption. Foundation for packaged offerings: creating repeatable solution bundles (templates, playbooks, and connectors) that can be deployed across multiple operators with configuration, not re-engineering. Integrating NOA into existing autonomous network solutions Most partners integrate NOA by keeping their domain platforms (assurance, orchestration, inventory, ticketing, observability) as the systems of record, while positioning NOA as the agentic orchestration and experience layer that coordinates people, tools, and workflows end-to-end. Practically, this integration typically follows a repeatable sequence: Choose the first high-value workflow (for example: P1 incident triage, RAN anomaly investigation, transport fault isolation, or trouble ticket enrichment) and define clear success metrics (MTTD/MTTR reduction, ticket quality, deflection rate). Connect tools through MCP and Open APIs, exposing partner and operator capabilities (queries, actions, validations) as governed tools the agents can call deterministically. Ground agents in partner/operator knowledge by connecting SOPs, topology, inventory, prior incidents, and KPI definitions via retrieval and curated context. Implement policy gates and RBAC so that high-risk actions (config changes, mass remediation, ticket closure) require explicit human approval and are fully logged. Operationalize with testing and observability, using evaluation runs, traces, and runbooks to validate correctness and monitor agent behavior in production. Where partners extend NOA to differentiate Partners extend NOA to differentiate on top of an open agentic foundation by combining Microsoft’s orchestration with deep domain expertise and proprietary IP. Packaged solutions, multi‑vendor interoperability, and a consistent Teams/Copilot‑based operations experience ensure scalability while preserving a familiar NOC workflow. Domain-specialized agents: adding RAN, core, transport, edge, and security agents tailored to vendor-specific telemetry, counters, and remediation procedures. Proprietary reasoning and models: incorporating partner algorithms (anomaly detection, RCA graphs, topology analytics) and selecting models suited to latency/cost/regulatory needs. Closed-loop automation: integrating with orchestration/controllers to move from “recommend” to “execute” in bounded, policy-approved scenarios (for example, self-healing with automatic rollback). Vertical solution packaging: delivering repeatable “packs” (connectors + prompts + workflows + dashboards) for specific operator pain points. Multi-vendor interoperability: normalizing data and actions across heterogeneous network domains using TM Forum Open APIs and partner mediation layers. Operations experience: embedding NOA into partner portals and NOC tooling while maintaining a consistent Teams/Copilot experience for daily operations. Co-innovation and operating model Successful partner implementations treat NOA as a living operations capability that is improved continuously—not a one-time integration. Partners typically establish a shared lifecycle with operators that covers solution design, governance, deployment, and ongoing optimization. Clear RACI: which actions agents may take autonomously vs. which require NOC supervisor approval vs. which must be escalated to engineering. Evaluation and release management: versioned prompts/workflows, pre-production evaluation runs, and rollback plans aligned to telco change-control practices. Safety-by-design: read-only defaults, least-privilege tool access, and explicit policy gates around configuration, customer impact, and data handling. Observability as a first-class requirement: tracing, action logging, and dashboards for quality, latency, and cost—plus incident reviews that include agent performance and tool outcomes. Knowledge ops: continuously curating SOPs, updating topology/inventory context, and incorporating learnings from resolved incidents to reduce repeat failures. Taken together, these adoption patterns show how NOA is becoming a common architectural baseline that partners can integrate into their portfolios and extend with domain expertise—helping operators move faster toward safe, scalable autonomy while preserving differentiation where it matters most. Partner Solutions Microsoft is actively working with a number of GSI and ISV partners on integrating NOA concepts into their solutions. Partners include: Global Systems Integrators Provide professional services to build, connect and ground you agent orchestration and data modernisation with Microsoft Fabric and Microsoft Foundry. Independent Software Vendors Turnkey, packaged solutions that include RAN & Core optimization agents, telco ontologies, or network insights which integrate or leverage the NOA architecture. Kenmei announced its collaboration with Microsoft to help operators accelerate their path toward autonomous networks by combining Kenmei’s telecom intelligence offer with Azure and Microsoft Fabric to enable scalable analytics and agentic AI–powered operations. Already in use at leading operators like Telefónica and etisalat (e&), this collaboration brings proven deployments into a broader cloud and AI ecosystem designed to reduce manual effort, speed decision‑making, and unlock new levels of network automation. Amdocs and Microsoft have further strengthened their strategic partnership by combining Amdocs’ aOS with Microsoft’s NOA Framework, amplifying the capabilities of the Amdocs Network Digital Twin. Leveraging Microsoft Fabric IQ’s unified data and intelligence platform, the Digital Twin evolves into a dynamic, AI-ready foundation that contextualizes inventory, assurance, analytics, and cross-domain data in real time. Together, the companies deliver intelligent network operations agents that detect anomalies, assess service impact, conduct root-cause analysis, and validate remediation scenarios prior to execution. This collaboration enables transparent, agent-driven automation that dismantles operational silos, improves efficiency, and accelerates CSPs’ advancement toward truly autonomous networks. Advancing the Journey Looking ahead, Microsoft remains committed to advancing telco autonomy through: Continued expansion of agent capabilities Future releases will bring additional agent roles, deeper coordination patterns, and broader integration with OSS/BSS systems—reflecting the trajectory of NetAI’s evolving multi‑agent ecosystem. Accelerated partner and customer enablement A downloadable NOA Framework accelerator, combining reference architectures, prompt libraries, agent templates, and deployment assets, will be made available in April 2026. Reinforcement of openness NOA will continue to support: Third-party agent onboarding Interoperability via MCP and TM Forum Open APIs Hybrid network environments Open‑source extensibility through Foundry components This ensures operators retain maximum flexibility while adopting autonomous operations. Closing: The Road to Autonomous Networks As telcos navigate increasing complexity and rising expectations, the enhanced NOA Framework provides a practical, modular, and secure path toward autonomous operations. By combining the intelligence of multi‑agent systems, the familiarity of Teams and Copilot, the power of Microsoft Foundry, and the safety of strong governance, NOA helps operators boost reliability, reduce MTTR, and simplify operations at scale. We invite you to explore the documentation, demos, and community resources at Mobile World Congress 2026 to learn how NOA can accelerate your autonomous network journey.
AI-Powered RAN and the Intelligent Edge: Microsoft’s Vision for the Future of Telecom
Artificial intelligence (AI) is rapidly converging with telecommunications infrastructure, promising to transform how networks are built, optimized, and monetized. Nowhere is this more evident than in the radio access network (RAN) – the crucial “last mile” that connects our devices to the digital world. At Mobile World Congress, Microsoft is sharing a strategic vision for AI in the RAN (AI-RAN) and intelligent edge computing. This vision centers on harnessing cloud and AI technologies to make telecom networks smarter, more efficient, and ready for new services. With decades of wireless research and a broad AI ecosystem spanning Azure to Copilot and Microsoft Foundry, Microsoft is partnering with the telecom industry to enable a new generation of AI-powered networks. A New Era of AI-RAN: AI Meets the Radio Network The concept of AI-RAN captures a threefold innovation in telecom networks, where AI and RAN technology intersect: AI for RAN: using advanced machine learning and AI algorithms to improve how RANs operate. By leveraging AI-based analytics and control, operators can dynamically optimize spectrum usage, network performance, and energy efficiency, leading to lower operational and capital expenditures. In practice, this means mobile networks that self-optimize – automatically adjusting parameters to reduce interference, enhance coverage, and cut power consumption without human intervention. AI on RAN: turning the RAN itself into a distributed AI computing engine. In this paradigm, the thousands of cell sites and edge data centers in a network can host AI inference workloads closer to end users. This intelligent edge approach allows telecom providers to offer new AI-driven services – from real-time translation to AR/VR and interactive gaming – with the ultra-low latency and data sovereignty that cloud alone cannot achieve. AI and RAN: creating a shared infrastructure where AI platforms and the RAN co-exist and collaborate. By co-locating AI resources with telecom network functions, operators unlock synergies like integrated sensing and communications (for example, using 5G cells as distributed sensors), and they can support “physical AI” use cases such as autonomous robots and smart factories at the edge. This convergence of AI and telecom infrastructure not only improves the network itself but also opens new revenue streams through innovative services delivered over 5G and future 6G networks. Microsoft envisions AI-infused RANs that are more than just communication channels – they become intelligent platforms for innovation. For instance, in recent trials, Microsoft researchers demonstrated AI systems that detect radio interference in real time by turning a 5G base station into a wideband spectrum analyzer. Similarly, an AI-driven anomaly detection system can continuously learn a network’s normal behavior and spot irregularities before they cause outages, helping prevent failures and improve reliability. These examples illustrate how applying AI to RAN data can translate into more resilient networks and better user experiences. Edge AI: Bringing Cloud Intelligence Closer The push for edge AI in telecom is about extending the power of the cloud out to the network’s edge, closer to where data is generated and consumed. This is crucial for applications that demand instantaneous processing and response, or that must keep data local for privacy and security. In a traditional setup, complex AI models live in the cloud, and lightweight AI runs on devices. But many new scenarios – such as unmanned aerial vehicle, autonomous mobile robot, or industrial IoT – require a middle ground. The telecom network’s edge (for example, in 5G base stations or nearby edge data centers) can serve as that ideal “in-between” AI execution layer. Edge AI offers several strategic advantages for operators and enterprises: Ultra-low latency: By processing data on edge servers just one “hop” away from end-users, critical applications (like autonomous driving or remote robotic control) can respond in milliseconds, far faster than sending data to distant cloud servers. Data sovereignty and privacy: Keeping sensitive data (such as video feeds, industrial sensor data, or health information) within local networks or on-premises helps meet regulatory and privacy requirements. AI at the edge can analyze data without that data ever leaving the telecom’s domain. Bandwidth optimization: By processing and filtering data locally, only the most important insights (or lightly compressed data) are sent to the cloud. This reduces backhaul traffic and lowers costs. Resilience and continuity: Edge AI systems can continue to operate even when connectivity to cloud is limited, ensuring critical services remain available. In short, intelligent edge computing transforms telecom networks into platforms for innovation. A prime example is the concept of “physical AI” – where AI-driven services control physical devices in real time via the network. Imagine factory robots or autonomous drones connected to a 5G network: with edge AI, heavy computation (like computer vision or coordination algorithms) can run on nearby servers, leveraging GPUs at the base station or aggregation site. Microsoft’s research has shown that offloading robotics AI workloads from onboard devices to edge GPUs can improve response times dramatically – in one scenario, cutting inference latency from over a second on a device to under 100 milliseconds at the edge. This kind of performance boost can make previously impossible applications feasible, from real-time hazard detection in smart cities to advanced augmented reality experiences. Unifying Cloud and Telecom through Microsoft’s AI Ecosystem Achieving the AI-RAN and edge vision requires more than just ideas – it demands a cohesive platform that brings cloud technology into the heart of telecom networks. This is where Microsoft’s broad AI and cloud ecosystem plays a pivotal role. Azure’s cloud platform provides the robust, scalable foundation. Telecom operators can run key network functions in Azure (such as 4G/5G core networks) and leverage Azure’s global infrastructure for high performance and elasticity. At the same time, Azure’s capabilities extend on-premises and to the edge via Azure Arc, enabling a single pane of glass for managing resources across public cloud, private data centers, and network edge sites. This means operators can deploy and manage AI models or applications on distributed RAN edge servers as easily as in the cloud – achieving “zero-touch” automation and unified operations across their entire network. Microsoft’s AI platforms and tools further empower telecom innovation. With Azure Machine Learning and the new Microsoft Foundry platform, operators and partners can train, fine-tune, and deploy state-of-the-art AI models for their unique needs. In fact, Microsoft’s AI ecosystem includes thousands of advanced models – from the latest OpenAI GPT-5.2 and domain-specific models, to a vast catalog of open-source models from partners like Anthropic, Meta, and Mistral – all available through Foundry for use in custom solutions. Likewise, Microsoft’s growing family of Copilot experiences and AI agent services can be harnessed to improve telecom operations and customer experiences. For example, the Network Operations Agent (NOA) Framework demonstrates how a service desk AI agent might assist network engineers by intelligently parsing through network alerts and suggesting fixes, while different agents could help automate customer support with industry-specific expertise. Under the hood, developers have access to powerful frameworks like the Semantic Kernel and Azure’s AI libraries to build their own telecom-focused AI applications and xApps (RAN applications) that run on cloud or edge infrastructure. Microsoft’s vision is to make developing AI-driven network solutions as seamless as any cloud application development – develop in Azure, deploy to the RAN. Crucially, all these capabilities are grounded in an open, standards-based approach. Microsoft is working closely with the industry to support Open RAN standards and has collaborated with leading operators and vendors on initiatives like Project Janus – an open RAN programmability platform that exposes rich RAN telemetry and control to AI algorithms. By embracing open interfaces and partnering across the telecom ecosystem, Microsoft ensures that AI solutions can plug into existing networks and equipment regardless of vendor, protecting operators’ investments while extending their capabilities. Microsoft is also a founding member of the global AI-RAN Alliance, a cross-industry effort to accelerate AI-native RAN technologies and establish best practices for integrating AI into next-generation networks. From Research to Reality: Innovation with Partners Microsoft’s leadership in AI and cloud is backed by deep research and real-world experimentation. Microsoft Research has been pushing the boundaries of wireless networking for over 20 years. Today, that research is yielding dividends in the form of new telecom technologies: Microsoft’s researchers have constructed a live AI-RAN testbed network across two global innovation hubs. This 24/7 private 5G network – spanning more than 30 cloud-controlled cell sites on Microsoft’s Redmond (USA) and Cambridge (UK) campuses – serves as a blueprint for the future RAN. It is fully software-defined, cloud-managed, and open, allowing internal teams to develop and test advanced 5G/6G capabilities like AI-driven optimization, edge robotics, and healthcare applications in a real-world environment. Insights from these efforts are shared with the industry and academy, helping define 6G-era concepts such as real-time RAN intelligent control and AI-native RAN architectures. Microsoft’s research prototypes (including reference designs and proofs-of-concept) offer operators a head start in understanding how to implement AI in their networks – from intelligent resource allocation to network slicing and beyond. Collaboration is key: Microsoft works hand-in-hand with major communication service providers (CSPs), network equipment manufacturers, and startups to bring these innovations to production. Joint trials and proof-of-concepts have demonstrated use cases like interference detection, energy-efficient RAN automation, and near-real-time network anomaly detection in live networks. By co-innovating with the telecom community, Microsoft ensures that its AI solutions align with real operational needs and can be deployed in multivendor environments. A Strategic Path Forward for the Telecom Industry As the telecom sector looks to the future, the message is clear: AI and the network are no longer separate – they are becoming one and the same. Operators that embrace AI-powered RAN and edge computing stand to benefit from significant gains in efficiency and customer experience. They will be able to optimize network performance in ways not possible before, from squeezing more capacity out of spectrum to slashing energy usage during off-peak hours. At the same time, these intelligent networks can unlock new revenue opportunities by offering differentiated services – think of carriers providing AI-powered insights or automation services to enterprise customers, or delivering rich digital experiences (from cloud gaming to mixed reality) with quality guaranteed by AI-driven network slices. Microsoft’s role is to serve as a platform and partner for this industry-wide transformation. By bringing its unparalleled cloud and AI ecosystem to the telecom domain, Microsoft is helping operators transform into hyperscale tech-driven enterprises. That means Azure infrastructure for carrier-grade reliability and scale, Azure ML and data platforms to train models on telecom data, Copilot and agent technologies to augment both network operations and customer-facing services, and the Foundry catalog of AI models and tools to jumpstart innovation. All of these building blocks are designed to work in a hybrid, open environment – spanning public and private clouds, the network core, and the far edge – so that AI can run wherever it creates the most value, even directly in the RAN. The convergence of AI and telecom infrastructure is poised to define the next decade of networks. Microsoft’s strategic investments in AI-RAN and edge computing, combined with deep partnerships across the telecom ecosystem, position it as a key enabler of this transformation. As the industry gathers at MWC to discuss what’s next, Microsoft reaffirms its commitment to helping telecom operators and partners harness the power of AI, from the cloud to the intelligent edge, and to jointly create a future where networks aren’t just faster or more open – but truly smarter.Introducing Microsoft’s Network Operations Agent – A Telco framework for Autonomous Networks
How NOA Works: Multi-Agent Intelligence with Human Oversight At its core, NOA is a multi-agent system tailored for telecom operations. It hosts a suite of specialized agents, each with a focused domain expertise – for example, one agent might handle network provisioning, another oversees software updates, and another focus on fault management. These agents continuously gather and interpret data from across the network and IT systems and feed their insights to a higher-level coordinating “planner” agent (NOA itself). The planner agent synthesizes inputs from all the specialists and generates real-time recommendations and insights for the operations team throughout the service lifecycle. In practice, this means many routine issues can be anticipated or resolved autonomously, with examples such as: Proactive deployment checks: During a new service rollout, a provisioning agent can automatically scan configuration scripts and flag anomalies or errors before they cause incidents, preventing outages caused by human error and improving overall network reliability. Accelerated incident response: If a network fault occurs, a service assurance agent springs into action to diagnose the issue. It can correlate telemetry and logs to pinpoint the root cause in seconds, then suggest the best remediation steps to engineers – massively reducing time to restore service. This shrinks the mean time to detect and repair issues, improving uptime. Crucially, NOA keeps humans in the loop. All agent-initiated actions operate under strict governance and operator-defined policies. Any automated fix or change recommended by an agent can be gated behind approvals, and every action is logged for audit compliance. This ensures that even as more tasks become automated, network engineers retain control and regulatory requirements are met. In short, NOA’s agents do the heavy lifting, but people set the guardrails. Key Components of the NOA Framework NOA brings together several Microsoft technologies into an integrated solution. Three foundational components make this telco agent framework powerful: Unified Data Access with Microsoft Fabric Effective AI agents require access to all relevant data, wherever it resides. NOA leverages Microsoft Fabric to break down data silos across the telco environment. Fabric acts as a unified data mesh for the network: it connects real-time telemetry streams, operational support system (OSS/BSS) databases, ticketing systems, and more into a single logical data layer. Broad data connectors: Fabric provides prebuilt connectors for Microsoft 365, Graph API, Dynamics 365, as well as telecom OSS/BSS and third-party systems. This means agents can directly tap into data ranging from network device metrics to customer trouble tickets, without custom integration work. Virtualized lakehouse (“OneLake”): Through OneLake, Fabric virtualizes multi-cloud and on-premises data into one scalable data lake. Whether the source is Azure Data Lake Storage, Amazon S3, Google Cloud Storage, or on-prem SQL servers, NOA’s agents can read and reason over it in real time without needing to physically relocate the data. Cross-domain data sharing: Fabric’s data virtualization and mirroring allow agents to combine insights across domains (e.g., correlating network performance data with service desk logs or even sales data) to make more informed decisions. By unifying telemetry and business data, NOA accelerates troubleshooting and decision-making. Agents and human analysts get a full picture of the network’s state and context instantly, improving accuracy of insights and enabling faster root-cause analysis. For the business, this means less downtime and more informed strategy, since decisions are based on comprehensive, up-to-date data. The framework is also aligned with industry standards like the TM Forum’s Autonomous Networks model, providing a common blueprint that fits into existing OSS/BSS processes. Microsoft has made available TM Forum–aligned templates, reference architectures, GitHub assets, and even Azure-hosted sandbox environments so that telcos can prototype and deploy their own agent-based solutions rapidly. Multi-Agent Orchestration with Azure Agent Framework A highlight of NOA is its multi-agent orchestration engine, built on the Azure Agent Framework. This open-source platform (part of Microsoft Foundry) provides the runtime environment and tooling to deploy, manage, and coordinate all the AI agents working in the system. In essence, it’s the “brain” that makes sure the right agent does the right task at the right time, and that they can communicate and work together seamlessly. Key capabilities of the Azure Agent Framework include: Standardized agent communication: Agents can talk to each other and to external services using open protocols. For example, Agent-to-Agent (A2A) messaging and the Model Context Protocol (MCP) allow dynamic tool use and data sharing between agents. This means a fault-management agent can trigger a troubleshooting agent automatically when needed, or an agent can call external APIs via OpenAPI definitions. Agent catalog and SDKs: Azure Agent Framework comes with a catalog of pre-built agent templates for common telco scenarios (provisioning, fault management, repair, etc.). Developers can also create custom agents using its SDK (with support for integration into existing apps), leveraging familiar tools like Visual Studio and GitHub for development and CI/CD. This drastically shortens the time to build new agents and integrate them into the NOA system. Built-in memory and observability: The framework provides long-term memory storage for agents and robust tracing/monitoring capabilities. This means agents “remember” past interactions and learn over time, and operations teams can monitor agent decisions and interactions in detail – crucial for refining agent behavior and troubleshooting any issues. It also includes enterprise-grade logging of agent actions (tying into the governance mentioned earlier). Enterprise security & hybrid readiness: Governance and security are baked in at the platform level. Agents can be deployed in a fully isolated manner (e.g. within Azure Virtual Networks), use managed identities for auth, and respect role-based access controls. The framework supports running agents in Azure or connecting to external/on-prem agent hosts, enabling hybrid and multi-cloud deployments out of the box. By using Azure Agent Framework, NOA ensures that a telco’s autonomous operations are running on a proven, secure, and extensible orchestration layer. (For more detail, see the Azure AI blog post “Introducing Microsoft Agent Framework” and the open-source Agent Framework repository on GitHub which provide deeper dives into these capabilities.) “UI for AI” – Copilot Integration in Teams and Outlook A distinguishing feature of Microsoft’s approach is making AI collaborative and user-friendly. Rather than confining insights to a dashboard, NOA integrates its agents into the tools where humans already work. This creates a “Copilot”-style experience for network operations. Through Microsoft Teams, Outlook, and the Copilot platform, NOA agents interact with engineers and managers in natural language: Conversational interface: An operations engineer can chat with the network AI agents as if they are teammates. For example, in a Teams channel, one could ask, “NOA, what’s causing the latency spike in region X?” and the agent would respond with its analysis, backed by data. Agents can also proactively post alerts or recommendations in chat when certain conditions are detected. Integrated into daily workflow: Within Outlook or Teams, if an incident occurs, an agent might automatically draft an incident summary or recommend next steps via a Copilot card, which the engineer can approve or tweak. This turns everyday collaboration tools into a unified operations cockpit where monitoring, troubleshooting, and decision-making happen collaboratively in real-time. Supervisor visibility and control: Managers can use the same interface to get high-level summaries, see trends (e.g., a weekly digest of recurring issues or network KPIs), and intervene when necessary. For instance, a supervisor could override an automated recommendation directly from within Teams if they see fit, or provide feedback to train the agents. With Microsoft 365 Copilot as the control system for these interactions, the learning curve is low – the AI fits into existing workflows. This “UI for AI” approach has proven to be a “killer app” internally at Microsoft: it dramatically improved productivity and response times in Microsoft’s own network operations by making human-AI collaboration seamless. The bottom line is that NOA’s advanced AI capabilities remain accessible and transparent to the people running the networks, rather than a black box. Open and Secure by Design The Network Operations Framework is built to be open and extensible. It’s not a closed system limited to Microsoft-only tools. Operators can integrate third-party or custom-built agents into NOA’s orchestration layer just as easily as first-party ones. For example, if a telecom has an existing AI solution or an OSS tool they want to include, they can wrap it as an agent and plug it into the framework. Microsoft’s AI Gateway service in Azure helps manage the security and identity of all agents (including third-party agents) via the MCP standard, ensuring consistent authentication, authorization, and compliance policies across the board. This open ecosystem approach means telcos can leverage their current investments and expertise, augmenting them with NOA, rather than starting from scratch. At the same time, NOA is secure by design. As mentioned, every agent action can require approval and is logged. The framework enforces read-only defaults for agents unless explicitly granted permissions. It uses restricted service accounts and integrates with existing access control systems (AAA/TACACS) to ensure agents only do what they’re permitted to do. Built-in guardrails prevent unsafe operations on network devices. This level of governance is critical in telecom environments, which are often highly regulated and sensitive. Automation is controlled – it operates within the bounds set by the network operators. Real-World Impact: Azure Networking’s Success Story Microsoft itself has been a “customer zero” for NOA, applying this framework to manage its vast global Azure network. The results demonstrate the transformative impact of autonomous operations. Microsoft’s Azure Networking team deployed multiple agents using the NOA framework to handle fiber optic incidents worldwide. These agents act as copilots and even fully autonomous responders for network fiber cuts and degradations – a traditionally labor-intensive domain. The outcome has been remarkable: Azure Networking achieved a 60% reduction in time-to-detect fiber issues and a 25% improvement in repair times. In other words, faults that used to take hours to even notice are now identified within minutes, and the restoration of service is significantly faster. Such improvements translate to higher network uptime and better customer experience. This example underscores how NOA’s combination of data-driven agents and automation can drastically improve operational efficiency in practice. Conclusion: A Blueprint for Telecom Autonomy The Microsoft Network Operations Agent Framework (NOA) offers telecom operators a pragmatic path to achieve autonomous networks. It’s modular, open, and built on proven technology – from AI agents and data fabric to collaboration tools – that operators may already use. Whether you are looking to modernize a Network Operations Center (NOC), automate fiber-optic repairs, or build a self-healing, self-optimizing network, NOA provides the foundation and tools to get started. It brings the promise of AI-driven autonomy within reach of network operators – augmenting human teams with intelligent agents to handle complexity at cloud scale. By adopting this framework, telecoms can improve reliability and performance today, while setting the stage for the fully autonomous networks of the future. Learn more: Read the NetAI White Paper https://aka.ms/netai_wpdl Check out the Microsoft Azure Blog announcement on the Agent Framework for the developer side of this technology. Explore the Agent Framework on GitHub to see how multi-agent systems are built. Read Microsoft’s Tech Community blog on Introducing Teams Mode for Microsoft 365 Copilot which illustrates the power of bringing agents into collaborative workflows. Review the Azure API Management AI Gateway documentation details how third-party AI agents can be securely managed in this ecosystem. With NOA, Microsoft is delivering a telco-specific blueprint for autonomous operations – and inviting the industry to build upon it.Reimagining Telco with Microsoft: AI, TM Forum ODA, and Developer Innovation
The telecom industry is undergoing a seismic shift—driven by AI, open digital architectures, and the urgent need for scalable, customer-centric innovation. At the heart of this transformation is TM Forum Innovate Americas 2025, a flagship event bringing together global leaders to reimagine the future of connectivity. Microsoft’s presence at this year’s event is both strategic and visionary. As a key partner in the telecom ecosystem, Microsoft is showcasing how its technologies—spanning AI, cloud, and developer tools—are enabling Communication Service Providers (CSPs) to modernize operations, accelerate innovation, and deliver exceptional customer experiences. 🔑 Key Themes Shaping the Conversation Connected Intelligence: Microsoft is championing a new model of collaboration—one where AI systems, teams, and technologies work together seamlessly to solve real-world problems. This approach breaks down silos and enables intelligent decision-making across the enterprise. AI-First Mindset: From network optimization to customer service, Microsoft is helping telcos embed AI into the fabric of their operations. The focus is on building shared data platforms, connected models, and orchestration frameworks that scale. Customer Experience & Efficiency: With rising expectations and increasing complexity, CSPs must deliver faster, smarter, and more personalized services. Microsoft’s solutions are designed to enhance agility, reduce friction, and elevate the end-user experience. As the event unfolds, Microsoft’s sessions and showcases will highlight how these themes come to life—through real-world implementations, collaborative frameworks, and developer-first tools. Thought Leadership & Sessions At TM Forum Innovate Americas 2025, Microsoft is not just showcasing technology—it’s sharing a bold vision for the future of telecom. Through a series of thought-provoking sessions led by industry experts, Microsoft is demonstrating how AI, open standards, and developer tools can converge to drive meaningful transformation across the telco ecosystem. From enabling intelligent collaboration through the Azure AI Foundry, to operationalizing AI and Open Digital Architecture (ODA) for autonomous networks, and empowering developers with GitHub Copilot, Microsoft’s contributions reflect a deep commitment to innovation, scalability, and interoperability. Each session offers a unique lens into how Microsoft is helping Communication Service Providers (CSPs) modernize their IT stacks, accelerate development, and deliver exceptional customer experiences. Microsoft Thought Leadership Sessions CASE STUDY: Connected Intelligence: multiplying AI value across the enterprise 📅Sep 10 1:30pm CDT Peter Huang, Senior Director, Technology, Network Data and AI T-Mobile Andres Gil, Industry Advisor/Business Developer, Telco, Media and Gaming Industry Microsoft CASE STUDY: From hype to impact: operationalizing AI in telco with TM Forum’s ODA and Open APIs 📅Sep 11 1:30pm CDT Puja Athale, Director - Telco Global Azure AI Lead Microsoft Connected Intelligence & Azure AI Foundry: Scaling AI Across the Telco Enterprise T-Mobile and Microsoft are spotlighting a transformative approach to enterprise AI: Connected Intelligence. The joint session explores how telcos can break down silos and unlock the full potential of AI by enabling strategic collaboration across systems, teams, and technologies. The core challenge they address is clear: AI in isolation cannot answer even the simplest customer questions. Whether it's billing, device performance, or network coverage, fragmented systems lead to blind spots, duplication, and poor customer outcomes. To overcome this, they propose a unified framework that blends technology and culture—because tech alone doesn’t scale, and culture alone doesn’t transform. Azure AI Foundry: The Engine Behind Connected Intelligence At the heart of this vision is Microsoft’s Azure AI Foundry, a shared AI platform designed to scale intelligence across the enterprise and a core component of Microsoft’s recently announced Network Operations Agent Framework. Connected Intelligence integrates: Agent Frameworks and Agent Catalogs for modular AI deployment Hundreds of TBs of daily data from network switches, device logs, and location records Enterprise-grade orchestration and data governance AI/ML models aligned with customer-level time series events This architecture enables reuse, speed, and alignment across people, organizations, and systems—turning data into actionable intelligence. Model Context Protocol (MCP): AI-to-AI Collaboration A standout innovation is the Model Context Protocol (MCP), which goes beyond traditional APIs. While APIs connect systems through data, MCP connects intelligence through context. It allows AI agents to dynamically discover and chain APIs without custom coding, enabling real-time collaboration across network operations, device management, and deployment workflows. By integrating MCP into the API fabric, Microsoft is laying the groundwork for agentic AI—where intelligent systems can autonomously interact, adapt, and scale across the telco ecosystem. From Hype to Impact: Operationalizing AI in Telco with TM Forum’s ODA and Open APIs The telecom industry is moving from hype to impact by operationalizing AI through TM Forum’s Open Digital Architecture (ODA) and Open APIs. The session, From hype to impact: operationalizing AI in telco with TM Forum’s ODA and Open APIs, explores how telcos can build AI-ready architectures, unlock data value for automation and AI agents, and scale responsibly with governance and ethics at the core. Microsoft’s collaboration with TM Forum is enabling telcos to modernize OSS/BSS systems using the ODA Canvas—a modular, cloud-native execution environment orchestrated with AI and powered by Microsoft Azure. This architecture supports plug-and-play integration of differentiated services, reduces integration costs by over 30%, and boosts developer productivity by more than 40% with GitHub Copilot. Learn how leading telcos like Telstra are scaling AI solutions such as “One Sentence Summary” and “Ask Telstra” across their contact centers and retail teams. These solutions, built on Azure AI Foundry, have delivered measurable impact: 90% of employees reported time savings and increased effectiveness, with a 20% reduction in follow-up contacts. Telstra’s success is underpinned by a modernized data ecosystem and strong governance frameworks that ensure ethical and secure AI deployment. From Chaos to Clarity with Observability Despite advances in operational tooling, fragmented observability remains a persistent challenge. Vendors often capture telemetry in incompatible formats, forcing operations teams to rely on improvised log aggregators and custom parsers that drive up costs and hinder rapid incident resolution. Microsoft’s latest contribution to the Open Digital Architecture (ODA) initiative directly tackles this issue with the ODA Observability Operator, now available as open source on GitHub. By enforcing a standardized logging contract, integrating seamlessly with Azure Monitor, and surfacing health metrics through TM Forum nonfunctional APIs, the operator streamlines telemetry across systems. Early trials have shown promising results—carriers significantly reduced the time needed to detect billing anomalies, enabling teams to shift from reactive troubleshooting to proactive optimization. Accelerating TM Forum Open API Development with GitHub Copilot As the telecom industry embraces open standards and modular architectures, Microsoft is empowering developers to move faster and smarter with GitHub Copilot—an AI-powered coding assistant that’s transforming how TM Forum (TMF) Open APIs are built and deployed. Why GitHub Copilot for TM Forum Open APIs? TMF Open APIs are a cornerstone of interoperability in telecom, offering over 100 standardized RESTful interfaces across domains like customer management, product catalog, and billing. But implementing these APIs can be time-consuming and repetitive. GitHub Copilot streamlines this process by: Autocompleting boilerplate code for TMF endpoints Suggesting API handlers and data models aligned with TMF specs Generating test plans and documentation Acting as an AI pair programmer that understands your code context This means developers can focus on business logic while Copilot handles the heavy lifting. Real-World Uses Telco developers benefit from powerful features in GitHub Copilot that streamline the development of TMF Open API services. One such feature is Agent Mode, which automates complex, multi-step tasks such as implementing TMF API flows, running tests, and correcting errors—saving developers significant time and effort. Another key capability is Copilot Chat, which provides conversational support directly within the IDE, helping developers debug code, validate against TMF specifications, and follow best practices with ease. Together, these tools enhance productivity and reduce friction in building compliant, scalable telecom solutions. For example, when building a Customer Management microservice using the TMF629 API, Copilot can suggest endpoint handlers, validate field names against the spec, and even help write README documentation or unit tests. 📈 Proven Productivity Gains CSPs like Proximus have reported significant productivity improvements using GitHub Copilot in their Network IT functions: 20–30% faster code writing 25–35% faster refactoring 80–90% improvement in documentation 40–50% gains in code compliance Other telcos like Vodafone, NOS, Orange, TELUS, and Lumen Technologies are also leveraging Copilot to accelerate innovation and reduce development friction. Best Practices for TMF API Projects To get the most out of Copilot: Use it for repetitive tasks and pattern recognition Always validate generated code against TMF specs Keep relevant spec files open to improve suggestion accuracy Use Copilot Chat for guidance on security, error handling, and optimization GitHub Copilot is more than a coding assistant—it’s a catalyst for telco transformation. By combining AI with TMF’s open standards, Microsoft is helping developers build faster, smarter, and more consistently across the telecom ecosystem. Learn more about how to configure and use GitHub Copilot in your own TMF Open API projects in our latest tech community blog. Microsoft’s Broader Vision for Telco Transformation Microsoft’s contributions reflect a comprehensive strategy to reshape the telecom landscape through scalable intelligence, open collaboration, and developer empowerment. At the core of Microsoft’s vision is the idea that AI must be connected, contextual, and reusable. The Azure AI Foundry and Model Context Protocol (MCP) exemplify this approach by enabling telcos to: Harness massive volumes of time-series data from networks, devices, and customer interactions Deploy modular AI agents that can collaborate across systems Orchestrate workflows that adapt in real time to changing conditions This architecture transforms fragmented data into actionable insights, allowing CSPs to move from reactive operations to proactive intelligence. Conclusion: Microsoft’s Strategic Alignment with TM Forum Microsoft’s participation at TM Forum Innovate Americas 2025 reflects a deep commitment to transforming the telecom industry through AI-first innovation, open collaboration, and developer empowerment. From T-Mobile’s vision for Connected Intelligence, to Microsoft’s roadmap for operationalizing AI and ODA, and the developer-centric acceleration enabled by GitHub Copilot, Microsoft is helping Communication Service Providers (CSPs) move faster, scale smarter, and deliver better customer experiences. By aligning with TM Forum’s goals—standardization, interoperability, and autonomous operations—Microsoft is not just participating in the conversation; it’s helping lead it. 📣 Call to Action Join Microsoft and other industry leaders at TM Forum Innovate Americas 2025 to explore the future of telco transformation. Whether you're a strategist, technologist, or developer, this is your opportunity to connect, learn, and shape what’s next.Unifying Data and AI in Telecom: Inside the Telco Analytics PoC Accelerator
As telecom operators race to modernize their operations and deliver personalized, data-driven services, the need for a unified, intelligent analytics platform has never been greater. The Telco Analytics PoC (TAP) Accelerator is Microsoft’s answer to this challenge: a deployable, open-source solution that brings together the power of Microsoft Fabric, Power BI, Azure AI, and Purview to help telcos unlock the full potential of their data. What Is the TAP Accelerator? The TAP Accelerator is a pre-packaged, cloud-native PoC environment that enables telcos to explore and showcase AI-powered analytics in action. It includes: A demo web app Power BI dashboards for tracking telco specific KPIs and reports Microsoft Fabric Lakehouses with a telco data model Sample telco data ML notebooks Real-time telemetry via Eventhouse Automation scripts and ARM templates for deployment It’s designed to be deployed in a customer’s Azure subscription, allowing hands-on customization and exploration of real-world telecom scenarios. Why It Matters Telcos face a unique set of challenges: Fragmented data across legacy and cloud systems Slow, manual decision-making processes Difficulty demonstrating the ROI of AI and analytics The TAP Accelerator addresses these by offering a unified, governed, and scalable platform that brings together data from across the business—network, finance, customer service, and sales—and applies AI to drive actionable insights. Core Capabilities Unified Data Integration Connects on-prem and cloud data sources via Microsoft Fabric and OneLake Uses a data mesh approach to unify domains without duplication Real-Time and Historical Analytics Eventhouse enables real-time telemetry (e.g., call center, network KPIs) Fabric Lakehouses support historical trend analysis and forecasting AI-Driven Insights Azure AI Foundry powers predictive models (e.g., churn, campaign ROI) Data Agents enable Copilot-style natural language querying Governance and Security Microsoft Purview ensures compliance, access control, and auditing Built-in security best practices for customer deployments Key Components Layer Description Fabric Lakehouses Bronze, Silver, and Gold layers for structured data processing Power BI Reports Dashboards for Call Center, Finance, Network, Sales, CEO views Eventhouse Real-time KQL-based telemetry ingestion and analytics Semantic Models Structured telco semantic models for Power BI and Copilot Data Agents Natural language interface for querying data Azure AI Services Cognitive services and ML models for advanced analytics Telco Data Model The TAP Accelerator uses the Azure Synapse database template for wireless that helps to redefine how data is managed and utilized within the telecommunications sector. This model is designed to streamline operations, foster innovation, and enable a more seamless integration of services across the industry. It's a comprehensive telco data model consisting of more than 20 domains, and thousands of tables. For example, the Network domain alone contains 399 tables defining every entity and attribute commonly used in network use cases. The Accelerator also includes guidance for modeling the required telco data schema in Synapse, and importing it into your Microsoft Fabric environment. Business Value The TAP Accelerator delivers measurable outcomes across the telco value chain: Customer Experience: Personalized engagement and faster issue resolution Operational Efficiency: Streamlined workflows and reduced overhead Financial Control: Real-time visibility into revenue, disputes, and liabilities Strategic Agility: Faster, AI-informed decision-making Security & Compliance: Enterprise-grade governance across all data flows Deployment and Access The solution is open-source and available on GitHub under a Microsoft license. It can be deployed via automation scripts into a customer's Azure and Fabric environment. Get started now!
Monetizing generative AI: How telecoms are unlocking new revenue streams
Introduction As someone deeply engaged in AI-driven transformations within the telecoms industry, I’ve witnessed firsthand how telecoms are spearheading the adoption of generative AI (GenAI). By leveraging AI-powered solutions and forming strategic alliances, telecoms are enabling enterprises to streamline operations, enhance customer experiences, and—most importantly—unlock new revenue opportunities. A recent survey of telecom executives underscores the growing momentum behind AI adoption, with nearly 50% of telcos reporting tangible benefits from GenAI—double the adoption rate from the previous year. Early adopters are already seeing meaningful returns, achieving cost efficiencies and enhancing customer engagement through AI-driven hyper-personalization. For example, one telecom refined its upselling techniques using GenAI, resulting in a 5–15% increase in average revenue per user (ARPU). Another deployed an AI-powered help desk bot, reducing per-call costs by 35% while increasing resolution rates by 60%. The urgency to monetize AI GenAI is becoming essential for tackling industry challenges such as streamlining operations and reducing costs through automation, accelerating growth via hyper-personalized marketing and customer insights, enhancing customer service through AI-driven virtual assistants and chatbots, and evolving telcos into “techcos” that deliver AI-driven services beyond basic connectivity. AI’s Impact on Business Strategies and Revenue Models Generative AI is reshaping the telecom landscape by optimizing pricing models, enabling personalized customer interactions, and reducing churn rates. AI-driven analytics can proactively identify customers at risk of switching providers, allowing telcos to deploy personalized retention strategies such as loyalty programs and customized pricing tiers. Modern monetization models require flexibility, allowing telcos to integrate a mix of subscription-based, usage-based, and one-time payment plans that cater to evolving customer demands. Monetizing AI in telecoms AI-Optimized Computing Services / GPU as a Service: With the demand for GPUs far exceeding supply, telcos can monetize their data centers by offering AI computing power to enterprises and government entities seeking sovereign AI solutions. AI-Driven Customer Engagement Platforms: Telecoms can package their AI-enhanced customer service capabilities as enterprise solutions for companies managing high-volume call centers. Intelligent Network Optimization: AI-powered Radio Access Network (RAN) solutions are improving network performance through real-time analytics, predictive maintenance, and dynamic resource allocation. Centralized AI Platforms / LLM as a Service: Leading organizations are developing centralized AI platforms that serve as repositories of proven and maintained AI/gen AI modules, APIs, tools, and code snippets. This platform approach helps drive quicker implementation of successful use cases while maintaining consistent guardrails and leveraging proven architectures and use-case “recipes.” For example, by building a GenAI platform with ~50 reusable services, one telecom successfully reduced the time it took to build new use cases from months to about two weeks. This ensured that all similar use cases used consistent architectures and that best practices and learnings were shared in a common repository. AI-powered fraud detection and risk management: AI can analyze vast volumes of transactional and behavioral data in real time to detect anomalies and prevent fraud. By offering fraud detection as a service or embedding it into enterprise solutions, telcos can reduce revenue leakage and enhance customer trust—both of which directly impact the bottom line. AI-enhanced personalized marketing: By leveraging customer data and behavioral insights, telecoms can use AI to deliver hyper-personalized offers, upsell opportunities, and loyalty programs. These targeted campaigns increase conversion rates and average revenue per user (ARPU), making marketing spend more efficient and profitable. AI-driven field operations optimization: AI can streamline field service operations by predicting equipment failures, optimizing technician dispatch, and automating maintenance workflows. These efficiencies reduce operational costs and improve service reliability—both of which contribute to margin expansion and customer satisfaction. Success stories: AI-driven transformation in telecoms Several telecoms are already seeing significant ROI from their AI investments. One achieved an in-year ROI of more than 2x from GenAI implementations, contributing to a multibillion-dollar cost-reduction target. Another reported an ROI of 9x to 12x from its proprietary AI tool, which optimizes customer support and internal workflows. A third launched a platform to democratize AI adoption, with real-world applications showcased at a major global event. High-ROI AI applications for telecoms AI is proving to be a game-changer across several key areas. In operational efficiency, AI-driven automation reduces workloads and enhances workforce productivity. In customer engagement, personalized AI-driven interactions improve satisfaction and retention rates. For network optimization, AI-powered analytics predict outages and enhance service reliability. In product innovation, AI enables more customized offerings, increasing customer loyalty. For fraud prevention, AI’s pattern-recognition capabilities enhance fraud detection and mitigate risks. In sustainability initiatives, AI helps telecoms minimize their carbon footprint by optimizing energy use and device recycling programs. Conclusion Generative AI is fundamentally reshaping the telecom industry, ushering in a new era of automation, intelligence, and monetization. By investing in robust AI frameworks and monetization strategies, telecoms can improve efficiency, expand revenue streams, and secure their leadership in the evolving digital economy.
