Case study · 01 · 2025 — Now
ALTO Studio
Enterprise AI platform bringing conversational AI, RAG, research, document workflows, and reusable APIs into one governed product.
Head of ALTO Studio R&D
PythonGoReactLangGraphKubernetesAzure OpenAIAWS Bedrock
Impact
- 6,500 users across Amundi
- 15+ production microservices
- 25-person engineering team
Context
- Amundi needed one governed entry point for AI instead of scattered team-level experiments with duplicated cost and risk.
- The platform had to serve very different audiences — analysts, portfolio managers, and product teams — in a regulated environment.
- Adoption, reliability, and cost control mattered as much as model quality from day one.
Approach
- Defined clear platform boundaries: gateway, knowledge services, application components, and operational tooling.
- Built conversational AI, RAG agents, DeepSearch, document extraction, and workflow automation as composable services.
- Organized the 25-engineer R&D team around architecture, delivery, hiring, and cross-team adoption.
- Shipped reusable APIs and components so product teams integrate AI instead of rebuilding it.
Lessons
- Enterprise AI platforms win on adoption and governance, not demos — stable APIs beat notebooks.
- Cost visibility by team and user is a feature, not an afterthought.
- Enablement (docs, patterns, support) determines whether a platform gets used.