Customer Service Automation · Enterprise Knowledge Search · IT Operations Automation
SAP needed to reduce repetitive support demand, break down knowledge silos, and automate business processes for regulated, multilingual customers without sacrificing European data localization or trust in AI-generated answers.
Reported impact
SAP's Mistral-powered framework cut repetitive support demand at SBB, scaled from a 1,000-user pilot to 30,000 employees, continuously expanded its knowledge base through expert feedback, and established a reusable path toward autonomous finance and procurement workflows for regulated industries.
SAP built a reusable enterprise AI framework on top of Mistral Large to support multilingual knowledge work and agentic business processes while keeping data hosted on SAP operated infrastructure in Europe. In the first deployment at Swiss Federal Railways, the team implemented a chatbot that uses retrieval-augmented generation to pull answers from FAQs, training materials, recordings, and screenshots, then applies glossary-based query rewriting so abbreviations such as "SO" are expanded before retrieval. When the bot cannot answer, human-in-the-loop escalation captures expert responses and feeds them back into the knowledge base. The service is surfaced inside SAP Build Work Zone and Microsoft Teams, and the same hosted framework is being extended to autonomous agents such as SAP's Accounting Accruals agent that reads SAP system data and policy documents to prepare draft journal entries.
FAQs, training materials, recordings, and screenshots feed a RAG knowledge base; employee questions arrive through SAP Build Work Zone or Microsoft Teams, glossary rewriting expands domain abbreviations before Mistral Large retrieves and answers, and unresolved cases are escalated to experts whose responses are written back into the knowledge base.
"80% reduction in repetitive support queries."
"The AI reads data from SAP systems and interprets accounting policy documents using Mistral's language understanding capabilities."
"The chatbot employs sophisticated retrieval-augmented generation (RAG) to draw from multiple knowledge sources including FAQs, training materials, and unstructured content like recordings and screenshots."
"The solution integrates directly into employees' existing workflows through SAP's Build Work Zone portal and Microsoft Teams"
"when the AI cannot resolve a query, expert responses are automatically captured and added to the knowledge base"
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
In-house Sample Mar 5, 2024 ServiceNow ServiceNow wanted to turn generative AI into a practical operating capability that could reduce support demand, speed agent work, and help developers complete software tasks more efficiently across internal digital… a 52% acceptance rate for generated code (text-to-code) Now Assist for Customer Service Management (CSM) · Now Assist for IT Service Management (ITSM) · Now Support 
Microsoft Sample Jan 13, 2026 Hertz Hertz needed to modernize hundreds of legacy processes—employee shift planning, roadside-assistance lookups across vehicle manuals on SharePoint, and payroll updates—to improve responsiveness for 26,000+ employees… 11,000+ rental locations in 160 countries Dataverse · Microsoft Copilot Studio · Microsoft Power Platform Microsoft Sample Nov 18, 2025 AT&T AT&T's employees were spending too much time investigating across many search links to find answers, developers were debugging code line by line, and data flowed faster than humans could keep up. Leadership needed proof… 100,000+ employees actively using agents Azure AI Foundry · Azure AI Search · Azure API Management 
