Swisscom, Switzerland's leading telecommunications provider with ~19,887 employees and CHF 11,036 million revenue, rebuilt its AI customer service agent Sam using Rasa's CALM framework — going from prototype to production in just 20 weeks, doubling automation rates and cutting operational costs by 50%. Writing and validation of conversation flows became 1.6x faster along the way.
"We aimed to create a more natural and human-like chatbot while still allowing handling complex interactions efficiently and according to business processes," said Rolf Neukom, Program Manager AI Service for B2C at Swisscom, framing the goal of the rebuild.
Sam originally launched in 2018 but had been constrained by an intent-based architecture that automated only a low two-digit percentage of customer interactions. Customers found the experience static and impersonal, while expanding Sam's functionality required time-consuming manual updates. In 2023, Swisscom partnered with Rasa to reimagine Sam as a dynamic AI assistant powered by generative AI, then worked over the following year to bring a new version of Sam to production.
The rebuilt agent blends LLMs with the reliability of business-process logic via Rasa's CALM framework. Zero-shot intent recognition lets Sam handle diverse queries without extensive manual training, and the agent can now pull real-time customer information to respond more effectively. Swisscom set clear goals on response time and per-session cost while real-time monitoring tracks key performance indicators and supports continuous improvement.
Beyond the headline numbers, Sam evolved from rigid intent-based responses to natural multi-turn conversations that handle follow-up questions, creating a substantially more engaging experience and lifting NPS. Optimizing LLM usage drove the 50% cost reduction, while automation freed human agents to focus on more complex work and faster deployment cycles let Swisscom adapt quickly to shifting priorities.
"Rasa was the perfect partner to bring our vision to life," Neukom said. The modular architecture is designed to extend Sam's functionality to voice channels and additional use cases, giving Swisscom a future-ready foundation for its customer service AI.
Details
- City
- Ittigen
- Organization
- Swisscom