NTT DOCOMO improves LLM usage analysis efficiency by 90% with Databricks
NTT DOCOMO built an internal 'LLM value-added platform' (approximately 10,000 monthly active users, 3 million monthly calls) and used the Databricks Data + AI Platform, including Model Serving for Azure OpenAI's GPT-4o and AI/BI Genie, to automate analysis of usage logs. This replaced manual Excel/Jupyter workflows, cutting monthly analysis work from 66 hours to 6 hours, a 90% reduction, while improving data governance via Unity Catalog.
Overview
NTT DOCOMO built an internal 'LLM value-added platform' (approximately 10,000 monthly active users, 3 million monthly calls) and used the Databricks Data + AI Platform, including Model Serving for Azure OpenAI's GPT-4o and AI/BI Genie, to automate analysis of usage logs. This replaced manual Excel/Jupyter workflows, cutting monthly analysis work from 66 hours to 6 hours, a 90% reduction, while improving data governance via Unity Catalog.
The challenge
As NTT DOCOMO's internal LLM value-added platform expanded, local tools like Excel and Jupyter Notebook could not handle the explosive growth in log data volume. Manual processes led to inaccuracies, duplication, and incomplete data, and storing logs locally on employees' development PCs through Excel and Jupyter Notebook posed data privacy and security risks since they often contained sensitive and confidential information.
The solution
NTT DOCOMO invested in the Databricks Data + AI Platform to conduct detailed analysis of LLM platform usage. The company uses Databricks Model Serving to deploy Azure OpenAI's GPT-4o model for log analysis, extracting and categorizing entities from user prompts and responses. Dashboards were built and deployed as production jobs using Databricks Lakeflow Jobs, automating tasks like updating, version control, and error detection. Fine-grained access controls were applied using Unity Catalog, and the company adopted Databricks' AI/BI Genie, a conversational analysis tool with a chat-style interface for natural-language data analysis and visualization.
Reported business value
NTT DOCOMO reduced the time spent manually processing and analyzing LLM usage data by 90%, cutting monthly analysis work from 66 hours to 6 hours. As of February 2025, the LLM value-added platform had approximately 10,000 monthly active users and 3 million monthly calls.
Sources
Open any source and check the claim yourself — that is the point of the register.
This record was researched and written with AI assistance, and its claims were checked against the sources above. (EU AI Act art. 50 transparency notice.)
Other telecommunications entries in the register.
AI serves QoS & operational efficiency
Orange France deployed AI-based tools across its network operations to improve quality of service and operational efficiency, including the Recovisu diagnostic tool that helps technicians correct errors during interventions, automated frequency declarations with the National Frequency Agency, AI-driven preventive maintenance for equipment such as antenna batteries, and an AI-evaluated citizen infrastructure-damage-reporting app. The NRJ5G app analyzed consumption data to prevent unnecessary mobile site electrical subscription upgrades, saving nearly 48 million.
Orange France· FranceRecovisu · NRJ5GDeutsche Telekom and Google Cloud Collaborate on MINDR Agentic AI for Autonomous Network Operations
Deutsche Telekom, in partnership with Google Cloud, announced MINDR (Multi-Agentic Intelligent Network Diagnostics & Remediation), a multi-agentic AI system built with Google Gemini models on Vertex AI and developed with Google Cloud's Autonomous Network Operations framework. MINDR correlates signals end-to-end across the RAN, transport, and core network domains to proactively identify service-impacting issues and support autonomous, explainable remediation, shifting operations from reactive troubleshooting to predictive automation. MINDR builds on Deutsche Telekom's RAN Guardian Agent, live in production in Germany since November 2025, which monitors network behavior during high-impact events and traffic surges. RAN Guardian autonomously triggered over 100 remediation actions at Christmas market events in its first month and reduced the time needed to manage major events from hours to around a minute (a more than 95% improvement). For 2026 it identified 237,000 events, including around 130 Carnival events each expected to draw over 10,000 participants, served by 611 mobile sites pre-checked for issues; only 5 sites experienced peak loads, which RAN Guardian optimized during the event. RAN Guardian is now scaling across Deutsche Telekom's European national companies, starting with the Czech Republic and Croatia. MINDR's first production releases are planned for later in 2026.
Chunghwa Telecom Builds Multiple Generative AI Solutions with Amazon Bedrock
Chunghwa Telecom, one of Taiwan's largest integrated telecom providers, migrated its generative AI projects to Amazon Bedrock and built three applications: a software development lifecycle documentation assistant, an interactive virtual English teacher using Amazon Titan and Stable Diffusion XL, and a generative AI marketing assistant built on Amazon Personalize.
BT Group writes and maintains 2 million lines of code a year using Amazon Q Developer
BT Group, a UK telecommunications company, rolled out Amazon Q Developer to around 2,000 active developers for AI-assisted coding, chat-based troubleshooting, automated unit test generation, and code transformation from older to newer, more secure versions.
Was this helpful?
Your feedback helps us improve our use case database