Azure Data Factory Migration and SQL Server Modernization for a Global Hospitality Platform

What We Did

• InterWorks delivered a phased, risk-managed approach to modernization.
• Discovery and scoping: The team began with a structured discovery phase to map every pipeline, understand its dependencies and validate the scope with business stakeholders. This became the foundation for detailed effort estimation and upfront risk management.
• SQL Server upgrade: As a deliberate stopgap and risk reduction step, the team upgraded SQL Server 2008 to 2017, decoupling that modernization from the larger orchestration migration and reducing variables.
• Architecture decision: The team evaluated multiple tooling options for the new orchestration layer. Early recommendations focused on Fivetran, but during discovery, the team pivoted to Azure Data Factory after recognizing that ADF's architecture better matched the complexity and interdependencies of SLH's specific workflows. This proved critical to the success of the delivery.
• Reverse engineering and documentation: The team reverse-engineered all 10 App Connect pipelines into textual representations, distilled the logic into FigJam module diagrams for particularly complex ones, and produced comprehensive documentation as a deliverable in parallel with the code. This documentation became the single source of truth for how each pipeline worked and why.
• AI-assisted implementation: The team used Claude to generate Azure Data Factory pipeline code from the textual reverse-engineered definitions. Rather than treating AI as a shortcut, the team used it as a productivity accelerator: AI generated the pipeline structure, InterWorks engineers reviewed, validated and iterated until the output matched the legacy behavior one-to-one. This approach kept the AI credit cost modest (roughly equivalent to three days of manual work) while compressing the timeline and producing fully version-controlled, documented code as a by-product.
• Validation and confidence building: Each pipeline was validated against the legacy system record by record, with business stakeholders directly comparing outputs to prove logical equivalence. This approach meant the organization could move to the new infrastructure with complete confidence that nothing had been lost or altered in translation.