Groupama Support & Service (G2S), the IT GIE for Groupama, is deploying generative artificial intelligence alongside Microsoft’s Power Platform and Copilot Studio to power conversational assistants like Robby, Raimond, and Sam. The integration aims to streamline document retrieval and administrative workflows across a large enterprise.
Scaling Enterprise Workflows with Low-Code and Copilot Studio
The convergence of generative AI and low-code architecture has fundamentally altered how enterprise applications are built. Rather than relying on traditional software development lifecycles for every minor tool, G2S has leveraged Microsoft’s Power Platform as a core engine for creating enterprise-grade applications. Since late 2024, Microsoft 365 Copilot has rolled out across the entire Groupama workforce, complementing custom-built internal assistants designed via Copilot Studio.
Arthur Halbaut, lead tech Power Platform at G2S, points to the integration of generative AI within Power Virtual Agent as the primary catalyst for this shift. The technology unlocked new operational capabilities, enabling conversational agents to parse massive document repositories, track administrative key performance indicators, and handle complex triage tasks like categorization, analysis, and quotation processing.
Inside Sam: Conversational AI Meets SharePoint
The practical application of this architecture is visible in tools like Sam. Embedded directly inside a custom PowerApp, Sam optimizes information retrieval across SharePoint repositories. Instead of deploying a standalone chat interface, the team built a complete search application designed around real user friction points.
If the AI agent fails to surface a relevant answer, users can either reformulate their queries or escalate the interaction to a human operator. Furthermore, if a user submits a negative rating on a generated response, the system automatically opens a tracking ticket inside ServiceNow. This closes the loop between automated generation and human oversight.
Every application, workflow, and AI integration within the Groupama information system must pass strict architectural reviews, compliance checks, and security validations. According to Arthur Halbaut, development follows a logarithmic trajectory:
This reality forces engineering teams to make definitive architectural choices once initial productivity gains level off. Clean, precise, and coherent documentation provided by business units remains an absolute prerequisite for maintaining agent efficacy.
Production Monitoring and the Shift Toward Agentic AI
To prevent model drift and ensure reliability, G2S established a rigorous monitoring framework for agents currently in production. Dedicated dashboards and evaluation suites run on a weekly cadence to verify that automated responses align with expected outputs. This continuous testing protocol gives business units clear visibility into actual usage metrics and functional performance.
Capitalizing on these initial deployments, G2S is setting its sights on advanced conversational systems and agentic workflows. Deploying autonomous agentic frameworks introduces unique structural hurdles within the heavily regulated insurance sector, where auditability and precision are non-negotiable.
G2S Enterprise Profile at a Glance
- Sector: Insurance
- Workforce: 32,000 employees
- Annual Revenue: 20 Md€
- Core Tooling: Microsoft Power Platform, Copilot Studio, SharePoint, ServiceNow
As G2S expands these capabilities, the intersection of low-code governance and generative models will dictate how legacy financial institutions scale automation without sacrificing security or operational transparency.
Related reading
