Industrialising AI on a group-wide Azure platform

Context

The group, which brings together several well-known insurance brands, had product teams experimenting with generative AI in parallel. Each pilot came with its own way of calling models, deploying code and handling data. To scale, the group needed one AI platform: shared components, automated delivery and the controls its risk teams expect.

Approach

Our team joined the platform engineering squad to build the components and services the platform exposes to product teams.

• Designed and delivered platform services on Azure Functions and Container Apps, exposed through Azure API Management.
• Automated the deployment and configuration of Azure resources with Terraform, and evolved the platform’s Azure DevOps pipelines.
• Supported product teams in industrialising their AI solutions: from notebook and prototype to a deployable, monitored service.
• Brought AI into engineering practices themselves, including agent tooling with MCP and LangGraph.
• Instrumented services with Dynatrace, Log Analytics and Application Insights.

Results

• One delivery path for AI services, from code to production, shared by product teams.
• Model access governed through the API management layer instead of direct calls.
• Product teams able to ship AI features on the platform without rebuilding the foundations each time.