{"slug":"capgemini-transforms-software-engineering-with-a-self-hosted-ai-assistant-powered-by-mistral","url":"https://findausecase.com/use-cases/capgemini-transforms-software-engineering-with-a-self-hosted-ai-assistant-powered-by-mistral","title":"Capgemini transforms software engineering with a self-hosted AI assistant powered by Mistral","description":"Capgemini built 'RAISE for Private Software Engineering', an internal, self-hosted AI coding assistant for regulated clients in aerospace, defense and public sectors who require on-premises AI. After switching the underlying model from an open-source LLM to Mistral's Codestral, without changing IDE or CI/CD integrations, code completion accuracy jumped from 50% to 90% and developer adoption rose from 30% to 100% across all client projects, with more than 50 client projects reporting productivity improvements within six months.","company":"Capgemini","industry":"Technology & Software","aiCapabilities":["Generative AI"],"businessFunctions":["IT & Software Engineering"],"technology":["Mistral Codestral","Mistral Code","Codestral Embed","Devstral"],"deployment":"On-Premise","problemStatement":"Capgemini's regulated clients in aerospace, defense, and public sectors had strict data-privacy and sovereignty requirements that ruled out cloud-based generative AI. Its earlier open-source coding assistant only reached 30% developer adoption, with ~50% accuracy that developers found insufficient for daily use, and it needed to support legacy applications dating to the 1980s across languages like ADA and C/C++.","solutionApproach":"Capgemini built 'RAISE for Private Software Engineering', a self-hosted internal software engineering assistant, and switched its underlying model from a previous open-source LLM to Mistral's Codestral without changing existing IDE or CI/CD integrations. Capgemini also invested in deployment methodology and developer coaching to drive adoption, and is now expanding to the full Mistral Code solution, including Codestral Embed and Devstral, for enhanced RAG and CI/CD integration.","businessValue":"Code completion accuracy jumped from 50% to 90%, developer adoption rose from 30% to 100% across all client projects, and more than 50 client projects reported productivity improvements in less than 6 months.","evidence":{"band":"high"},"sourceUrl":"https://mistral.ai/customers/capgemini/","dates":{"publishedAt":"2026-09-27T05:49:16.311Z","publishedAtSource":"pipeline","updatedAt":"2026-09-27T05:49:16.311Z"},"license":"Open for reading and citing with a link to https://findausecase.com/use-cases/capgemini-transforms-software-engineering-with-a-self-hosted-ai-assistant-powered-by-mistral. Bulk republication requires permission."}