← Cases
SwitzerlandLogistics and transport2025 – ongoing

SBB (Swiss Federal Railways)

SBB Chat, an LLM customer-service chatbot

SBB's new LLM chatbot resolved about 60% of customer questions in its first weeks, the railway said

WorkingCompany's own claim

Problem

Rail customers ask about a wide range of tickets, fares and connections, and the previous help channels left many questions to human agents.

Approach

SBB Chat is built on OpenAI's language model and adapted to SBB's own data with retrieval-augmented generation. It went live a few weeks before the July 2025 report. Questions it cannot solve go to a human agent.

Results

~60%share of questions the chatbot resolved without a human, as stated by SBB to SRFCompany's own claim[1]

Takeaway

SRF attributed SBB's lower resolution rate mainly to the complexity of rail services. Neon reported roughly 90% for its own help bot.

Categories

Business function
Customer service
Innovation type
Process innovation
Evidence
Company's own claim

Sources

Source summaries

[1]SRF8 Jul 2025

The Swiss broadcaster compares two new customer-service chatbots. SBB Chat, built on OpenAI's model with retrieval-augmented generation, resolves about 60% of questions according to SBB, with the rest passed to agents. Neon's bot, built on Google's Gemini, reports about 90% success.

Full details are in the original sources.

Added 25 Oct 2026 · Updated 25 Oct 2026
Open case page ↗