IT Operations Automation · Real-time Analytics
Deutsche Telekom needed a faster way to detect, classify, and resolve mobile network issues around demand spikes and other exceptional situations because growing network complexity made traditional automation insufficient for real-time problem solving.
Reported impact
Deutsche Telekom reports faster troubleshooting and optimization, more resilient network operations, and a better mobile customer experience, while positioning the implementation as a first step toward autonomous, self-healing networks.
Deutsche Telekom built the RAN Guardian Agent as a multi-agent AI system for mobile network operations using Google Cloud technology. The implementation starts by ingesting publicly available event signals from online directories and social media to identify upcoming events across Germany, estimate their scale, determine location, and automatically verify the collected information. A second stage combines those event inputs with network monitoring, inventory, performance, and coverage data to assess whether nearby antennas can handle expected traffic, monitor key parameters in real time, and classify degradations with Gemini 2.5-based reasoning. CloudRun is part of the runtime stack for the agent services, while BigQuery and Firestore are named components in the architecture. A third agent recommends and executes corrective actions such as reallocating resources or adjusting configurations, then documents actions and outcomes so network teams can improve readiness for future events.
The system gathers and verifies event signals from online directories and social media, then combines them with network monitoring, inventory, performance, and coverage data to assess expected traffic near each event. The agent stack uses Gemini 2.5 with CloudRun, BigQuery, and Firestore in the implementation to analyze antenna capacity in real time, trigger or recommend corrective actions, and document outcomes for future optimization and operator review.
"In the first step, an AI agent continuously scans and analyzes publicly available sources, such as online directories and social media, to identify upcoming public events across Germany."
"In the second step, another component of the RAN Guardian uses this data to assess how well the mobile network can handle the expected traffic during each event."
"In the third step, an additional AI agent executes these measures, for example by reallocating network resources or adjusting configurations. It also documents all actions and outcomes to continuously improve performance and enhance readiness for future events."
"It evaluates the capacity of nearby antennas, monitors key network parameters in real time, and can immediately recognize high utilization levels, recommending suitable optimization measures if required."
"Processes that previously took roughly an hour can now be completed within just a few minutes."
"The RAN Guardian Agent was created with Google's Gemini 2.5. In addition, it also uses CloudRun, BigQuery, and Firestore."
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
Google Cloud Sample Archived · Oct 16, 2026 Alta Alta needed a transactable Google Cloud Marketplace path for its AI and technology spend management platform without pulling its product team off core forecasting work, while also helping finance and IT leaders… Saved 125 engineering and 60 project management hours Cloud Pub/Sub · Cloud Run · Cloud SQL Google Cloud Sample Modified · Oct 9, 2026 Anthropic During large-scale training runs, traffic is highly variable—dropping to near zero during model checkpointing before bursting to millions of events per minute. Anthropic needed a serverless messaging queue that scaled… Handles peak throughput of over 1M events per second BigQuery · Bigtable · Cloud Pub/Sub Google Cloud Sample Archived · Oct 2, 2026 M Lhuillier The company needed to move away from legacy on-premise hardware and previous cloud environments that suffered from high maintenance costs, scalability bottlenecks during peak holiday remittance periods, and slow… Accelerated 2-year cloud transition into an 8-month rollout BigQuery · Cloud Run · Gemini 