Microsoft and Tech Mahindra launch Agentic AI platform to overhaul telecom data systems

The joint effort leverages Microsoft’s Azure AI services and Tech Mahindra’s telco‑grade data fabric to embed autonomous decision‑making directly into network operations, according to the launch announcement from Tech Mahindra and Microsoft. The platform, dubbed “Agentic AI,” is built to ingest raw telemetry from radio access networks, core routing elements, and subscriber‑level usage logs, then apply large‑language‑model (LLM) inference and reinforcement‑learning policies that trigger configuration changes without human intervention. By chaining data ingestion, model inference, and actuation pipelines, the solution promises end‑to‑end automation that can, for example, reroute traffic in milliseconds when congestion is detected or provision new virtual network functions on demand.

Tech Mahindra’s engineering lead described the system’s “real‑time analytics” layer as a continuous stream processing engine that normalizes heterogeneous data formats across legacy OSS/BSS stacks, then feeds them into Microsoft’s Azure OpenAI Service for contextual reasoning. The architecture relies on Azure’s confidential compute enclaves to protect sensitive subscriber information while still allowing the LLM to generate actionable insights, the partners said. In practice, the platform can surface anomalies such as sudden spikes in dropped calls, automatically generate remediation playbooks, and push those playbooks to network elements via standardized NETCONF/YANG interfaces—all without a manual ticket.

The announcement arrives amid heightened scrutiny of Microsoft’s AI security posture. Recent reporting by Forbes highlighted a sophisticated multi‑stage AI‑driven phishing campaign that exploited Microsoft 365 email accounts, while The Register noted that AI‑generated phishing attempts are now 4.5 times more effective, according to Microsoft’s own data. Reuters also reported that state‑backed actors from China, Russia and Iran have been using tools derived from Microsoft‑backed OpenAI models to refine their attacks. Although the Agentic AI platform is positioned for operational efficiency, the partners acknowledge that embedding LLMs in critical infrastructure raises new attack surfaces, prompting them to incorporate Microsoft’s Zero Trust and Azure Sentinel monitoring frameworks as part of the deployment kit.

Analysts familiar with the telecom sector see the move as a response to the industry’s “data‑deluge” problem, where traditional rule‑based automation struggles to keep pace with 5G and edge‑computing workloads. By offloading pattern recognition and decision logic to a centralized AI engine, operators can theoretically reduce OPEX tied to manual network optimization. However, the lack of disclosed performance benchmarks in the launch brief leaves open questions about latency, model drift, and the cost of continuous model retraining at scale. As the platform rolls out to pilot customers later this year, its real‑world impact will hinge on how effectively it balances the promised automation gains against the operational overhead of securing and maintaining an AI‑centric network stack.