DHL — RMA LOGISTICS: HYBRID DATA & AI PLATFORM

The situation

RMA Logistics, a regional hub for DHL, needed to bridge on-premises data systems and cloud infrastructure. Operational data was spread across Excel files, local databases and siloed departmental systems. Multiple regional distribution centres needed a shared view of that data.

The consequences were concrete. Invoice processing took 2 days of manual work, and operational insights could require days of data compilation. We designed and implemented a hybrid Azure Databricks platform to bring the sources together and automate data-intensive processes.

Key decisions

We chose a hybrid architecture to connect the existing on-premises systems with the cloud platform. VPN tunnels, ExpressRoute and private endpoints provided secure connectivity. This addressed the boundary between local systems and cloud infrastructure rather than treating data consolidation as a cloud-only task.

We built a unified Databricks Lakehouse across the distribution centres. Consolidating Excel files, local databases and departmental sources created a single source of truth for operational reporting and analytics. The shared data foundation supported both business-user access and predictive workloads.

For process automation, we focused on specific tasks: extracting and classifying invoice documents, reconciling financial data, and producing daily/weekly/monthly operational reports. Invoice processing fell from 2 days of manual work to under 10 seconds — a 99.5% time reduction.

We also established Azure DevOps CI/CD pipelines and Terraform infrastructure as code for automated deployment and hybrid infrastructure management. Our Azure Terraform practices provide related guidance for teams managing data-platform infrastructure.

What was hard

The central challenge was fragmentation across both systems and processes. Connecting infrastructure alone would not replace the manual Excel work. Consolidating data alone would not automate invoice handling or financial reconciliation. Our work addressed both: secure hybrid connectivity and a shared Lakehouse, alongside automation of the workflows that consumed the data.

Access to insights also needed to change. We implemented real-time performance dashboards and self-service analytics so business users could access data independently. Anomaly detection and models for demand forecasting and shipment volume prediction extended the platform beyond reporting, enabling proactive resource allocation across distribution centres.

Lessons for similar teams

Treat hybrid connectivity, data consolidation and process automation as connected parts of the same problem. In this engagement, each addressed a different source of manual work or fragmented information.

Be specific about what automation does. Document extraction, classification and reconciliation describe the work more clearly than a broad AI label. Likewise, distinguish reporting from prediction: real-time dashboards supported operational visibility, while forecasting supported resource allocation. Both depended on bringing scattered sources into a shared data platform.