← Cases
SwitzerlandTechnology and telecom2023 – ongoing

Swisscom

Rasa-based Sam customer-service assistant

Swisscom halved Sam's operational costs after rebuilding the assistant with Rasa, the vendor reported

WorkingVendor case studyVendor ↗

Problem

Swisscom's earlier Sam assistant struggled with follow-up questions and scaling. Its intent-based models automated only a limited share of customer interactions.

Approach

Swisscom began working with Rasa in 2023 to rebuild Sam using its CALM framework. The assistant combined large language models with defined conversation flows and access to customer information. The team moved from prototype to production in 20 weeks.

Results

50%reduction in operational costs for Sam, per RasaVendor case study[1]
Doubledautomation rate after rebuilding Sam, per RasaVendor case study[1]

Takeaway

Swisscom tracked response time and session cost during development. The reported savings concerned the assistant's operations, rather than total customer-service spending.

Categories

Business function
Customer service
Innovation type
Process innovation
Evidence
Vendor case study

Sources

Source summaries

[1]Rasa2026

Rasa described rebuilding Swisscom's Sam assistant with language models and controlled conversation flows. It reported doubled automation and a 50% reduction in operational costs.

Full details are in the original sources.

Added 1 Nov 2026 · Updated 1 Nov 2026
Open case page ↗