Snowflake laid out three predictive uses for SAP data with no copy. The outline from Eric Moeller centers on SAP Business Data Cloud Connect for Snowflake. Models run inside Snowflake governance.
Before the Zero-Copy Integration launch, extracts and rebuilds stalled most pilots.
Now SAP semantics arrive in Snowflake in near real time. Supply chain models join SAP sales with IoT readings, logistics data and weather. Marketplace adds weather, risk indices and macro data with no extra integration. Finance models join ledger data with CRM, vendor history and market signals. Cortex AI flags invoices outside normal ranges and scores risk. Churn models join S/4HANA, CRM and usage telemetry for next best action.
Connection starts with SAP BDC Connect for Snowflake and no ETL pipelines. Enrollment uses an invitation link generated by SAP BDC, per SAP community setup guide. SAP objects arrive with semantics like Customer, Product and General Ledger intact. Teams model in Snowpark or Cortex AI with no data movement. LLMs from Anthropic and OpenAI come to the data, not the reverse.
The claim is simple: keep semantics, add context, predict better. Predictions must publish back to SAP BDC and change shipments, payments and offers.





