Event Backbone
Kafka on Azure decoupling warehouses, TMS and portals from the ERP.

A European third-party logistics group ran 14 ERP instances inherited from acquisitions. We consolidated them into one S/4HANA core with an event-driven integration layer and an AI-assisted data migration, cutting the order-to-cash cycle by 38%.

Overview
After a decade of acquisitions the group operated 14 ERP instances in six countries, each with its own master data, pricing logic and reporting. Month-end closing took three weeks and cross-border orders were re-keyed by hand.
Over twelve months we led the consolidation onto a single S/4HANA core, built the integration layer that connects warehouses, transport management and customer portals, and used LLM-assisted mapping to migrate and reconcile master data from the legacy systems.
The Challenge
The same customer existed up to fourteen times with different addresses, terms and credit limits. Warehouse and transport systems depended on local ERP quirks nobody had documented.
The business could not stop. Every country had to keep shipping during the migration, and finance needed consolidated reporting from month one of the new core.

Our Solution
We introduced a Kafka event backbone on Azure first, so warehouses, TMS and portals talked to a stable contract instead of to a specific ERP. Countries then moved to the S/4HANA core one at a time, with the legacy instance kept in read-only mode for reconciliation.
Master data was the hardest part. An LLM-assisted mapping pipeline proposed matches and normalisations for customers, materials and price conditions; data stewards approved them in a review tool, and every decision was logged. The approach reduced manual reconciliation effort by roughly two thirds.

Key Features
Kafka on Azure decoupling warehouses, TMS and portals from the ERP.
Six migrations with no shipping interruption.
LLM-proposed matches reviewed by data stewards with full audit.
Group finance view from the first month on the new core.
Customer and dealer portals connected through stable APIs.
Documented operations for the group's internal SAP team.
The Impact
One core, one customer record, one set of numbers.
-38%
Order-to-cash cycle
14 → 1
ERP instances
6
Countries migrated
12 months
Programme duration