Deutsche Telekom, a global leader in telecommunications, faced significant challenges in modernizing its internal IT support system.
Routine inquiries overwhelmed the phone hotline, diverting resources from addressing complex issues. Legacy systems lacked the scalability and flexibility required to support a rapidly evolving workforce, while traditional NLU-based methods demanded substantial manual effort to maintain intents and entities.
Key Takeaways
- Deutsche Telekom needed a solution that could deliver both German and English support
- The RASA Platform empowered non-technical teams, such as IT service desk subject matter experts, to design and refine intricate conversational flows, enabling faster iteration and reducing the reliance on technical staff for every update
- CALM helped resolve 50% of inquiries autonomously and reduce agent workloads by 30%
Additionally, the need to support German and English-speaking employees highlighted the need for a strong multilingual solution to deliver consistent experiences across languages. The agent's limited ability to integrate with backend systems hindered its potential to automate workflows and provide personalized, efficient support.
Challenges
To modernize its support system, it turned to Rasa and its CALM (Conversational AI with Language Models) framework, which combines the flexibility and fluency of large language models (LLMs) with the precision and control of structured business logic.
A customized, bilingual solution fit for purpose
Deutsche Telekom partnered with Rasa to develop an advanced conversational AI agent solution powered by CALM to meet the organization’s unique IT support needs.
Rasa Studio empowered non-technical teams, such as IT service desk subject matter experts, to design and refine intricate conversational flows, enabling faster iteration and reducing the reliance on technical staff for every update. The agent accessed user specific data by integrating Rasa’s action server with Deutsche Telekom’s internal REST API gateway, enabling efficient interactions.
Made for Deutsche Telekom’s internal teams, the agent addresses workplace IT issues, offering employees a faster and more effective alternative to traditional phone hotlines. With bilingual support in German and English, the agent provides 24/7 assistance, delivering immediate resolutions to basic queries while escalating complex cases to human agents when needed.
via personal devices
RAG solution for precise query handling
dialogue creation with Rasa Studio
Precise results, contextual responses
The solution incorporated Deutsche Telekom’s Neural Information Retrieval (NIR) engine with a retrieval-augmented generation (RAG) approach, allowing the agent to deliver precise, contextually relevant responses to “How-Do-I” inquiries. The agent also introduced proactive support features, such as opening incident tickets, involving human agents in active chat discussions for unresolved issues, and operating on personal devices without requiring internet access. CALM’s modular architecture allowed continuous refinement, ensuring the solution adapted to real-world scenarios while maintaining robust performance and reliability.
CALM allowed Deutsche Telekom to deliver a scalable, secure, and user-centered digital assistant that addresses technical and employee needs.
"Rasa CALM is a product that helps you build chatbots with complex
dialogue structures and integrated backend actions easily and quickly,
making it more advanced than RAG-based chatbot solutions." - Slavomir Lihan, Value Stream Lead, Deutsche Telekom
Results
costs significantly
The agent resolved 50% of inquiries autonomously, reducing agent workloads by 30% and significantly lowering personnel costs. Employees benefited from immediate responses to routine queries, avoiding hotline wait times and accelerating resolution speeds. Its multilingual capabilities ensured consistent and effective communication in both German and English, enhancing accessibility for diverse employees.
Proactive support features further streamlined workflows, enabling the agent to address issues on personal devices. By shifting conversational design responsibilities to domain experts using Rasa Studio, developers and data scientists concentrated on strategic projects. The agent is widely used across the company, assisting 10K+ employees whenever they encounter IT and workplace-related issues.
Rasa CALM transformed Deutsche Telekom’s approach to IT support, delivering measurable improvements across multiple areas. Deutsche Telekom’s implementation of Rasa CALM redefined its IT support strategy. It delivered scalable, efficient, and multilingual conversational AI that elevated the employee experience across the organization.
By integrating advanced LLM capabilities with structured workflows, Deutsche Telekom created a scalable, efficient conversational AI agent for employee needs.
Looking ahead, Telekom plans to enhance the agent's dialogue quality, expand its reach to additional regions, and fine-tune custom language models to support more advanced use cases. Telekom has set a new benchmark for internal IT support with Rasa, proving that innovation, efficiency, and user-centric design coexist in enterprise AI solutions.
