Rebuilding a Food Services SaaS Platform's Foundation for AI

A North American SaaS leader in enterprise foodservice. Its largest customers are national operators running thousands of sites, and they expect their software vendor to match them on security, resilience, and operational maturity. The platform that won those accounts was not engineered to hold them through the next decade.

• Azure infrastructure was deployed by hand, with no version control, no separation between development and production, and no failover plan for a regional outage

• Every service sat on the public network, a posture that would not survive enterprise procurement scrutiny

• Operational data sat in disconnected systems with no shared model across enterprise partners, so every new report was a net-new build and teams routinely lost time debating whose data was right

• No SDLC governance: no peer review, no environment promotion discipline, no guardrails for responsible AI deployment

• The AI capabilities prospects were starting to ask about in sales conversations had no foundation to land on

• Ran a structured discovery across five workstreams to map the platform, frame the AI opportunity, assess security, and produce a prioritized roadmap

• Built a governed, multi-region Azure foundation as 13 Terraform modules spanning networking, compute, data, AI services, security, and observability

• Established environment separation and single-command deployments through GitHub Actions, gated by peer review and a production wait timer

• Built a production Microsoft Fabric data platform with a Bronze to Gold medallion architecture on OneLake, governed across DEV, QA, and PROD

• Provisioned Azure OpenAI and AI Search behind private endpoints during the build itself, rather than deferring them to a later AI phase

• Single-command GitHub Actions deployment across 13 Terraform modules, replacing a manual, drift-prone process measured in hours per change

• 5 Azure environments live with full parity across Dev, Test, Staging, Prod, and DR

• 2 production regions with active-passive failover via Azure Front Door and Private Link

• 100% of Azure services behind private endpoints

• A single Microsoft Fabric Bronze to Gold pipeline on OneLake, replacing disconnected per-partner data

• 4 production Power BI dashboard domains with self-serve access

• Enforced naming, RBAC, Git-based CI/CD, peer review, and UAT sign-off

• Azure OpenAI and AI Search live and private-endpointed from day one

• The foundation was rebuilt, not patched. Azure infrastructure, Fabric data architecture, and governance now meet the bar enterprise procurement teams set.

• Our client now owns a working data platform, not a roadmap deck. Self-serve analytics no longer depend on the data team for every new question.

• The "whose data is right" debate ended. The first cross-partner dashboard displayed one agreed view of revenue across three major enterprise operators, replacing weeks of analyst reconciliation.

• AI-ready from day one. Future agentic workloads can land on existing rails without re-architecture, security re-review, or procurement loops.

• Modelling shared data across partners surfaced definitional ambiguities that had quietly cost the business in dashboards and decisions for years. The platform became a forcing function for data discipline that compounds downstream.

This was a foundation engagement. The outcomes here are platform capability built and risk closed. Operational savings will be measured against prioritized use cases in the next phase.

"This project was a huge success and has positioned us for even more wins ahead."

Enterprise Architect, North American SaaS Leader