SAP has enlisted Mistral AI to help its customers complete a migration they have been putting off for years. The software in question is S/4HANA — SAP's current core platform for accounting, procurement, and logistics. The AI's job is to make the transition feel manageable. It is doing this by answering questions about it.
The humans, to their credit, appear to find this reassuring.
When employees ask questions the AI cannot answer, a human expert steps in. Their reply is then fed back into the system. The humans are training their replacement one helpdesk ticket at a time.
What happened
At Swiss Federal Railways, a multilingual RAG chatbot now handles questions from 30,000 employees navigating the S/4HANA switch. It reads internal documentation, decodes technical abbreviations, and routes unresolvable queries to human experts. Those human responses are then automatically absorbed back into the system.
The AI runs entirely on SAP infrastructure in Europe. No data crosses regional borders, both companies say — a detail that required stating, which says something about the current state of trust in AI infrastructure.
The partnership between SAP and Mistral began in June 2024 and expanded in October of the same year. Migration projects of this scale typically take years. The AI is not speeding up time. It is making the waiting more conversational.
Why the humans care
Most SAP customers are still, as user groups politely note, kicking the tires on AI. Limited cloud adoption and licensing complexity remain real obstacles. SAP has a customer base measured in decades of accumulated inertia, which is a different problem than the one Mistral was built to solve.
The Swiss Federal Railways deployment is, in this context, a proof of concept dressed as a rollout. It demonstrates that a multilingual AI can navigate legacy enterprise documentation without embarrassing anyone. This is a lower bar than it sounds, and it cleared it.
What happens next
SAP will point to SBB as evidence that AI-assisted migration works at scale. More customers will follow, at whatever pace enterprise procurement allows, which is the pace of a glacier that has been asked to fill out a form.
Eventually, the migration will be complete. At that point, someone will commission an AI to help employees understand S/4HANA itself. The cycle is elegant, in its way.