There Is No Box in SAP's Launchpad for Your Azure OpenAI Key
We're two PhD students behind SAP AI Lab — one of us does ML systems, the other enterprise information systems; independent blog, endorsed by neither vendor. Wiring Azure OpenAI into SAP BTP is the most common "first AI project" for SAP architects in 2026, and we watched two attempts fail the same way before we got it right ourselves. Verified as of 2026-11-08. Both teams started with the same…
Two PhD students from SAP AI Lab have found that there is no box in SAP's Launchpad for integrating Azure OpenAI keys. Both teams encountered the same issue when trying to register their Azure OpenAI endpoint in the SAP AI Launchpad. The SAP AI Core offers Azure OpenAI models through the azure-openai executable, running on SAP-managed Azure capacity.
This path, known as Path A, requires the AI Core extended plan and is billed through the SAP AI Core subscription. It is important to note that this path is gated by the fact that foundation-models and orchestration scenarios are not available in the Free or Standard service plans.
Path B involves bringing your own Azure OpenAI deployment and calling it directly from your BTP application. The API key is used, but a data-plane role, Cognitive Services OpenAI User, must be assigned to permit inference. This path requires a REST call from your BTP application straight to your Azure OpenAI deployment, with the API key working but requiring the data-plane role for successful inference.
An often overlooked step is the use of an orchestration layer, which provides a pipeline for grounding, templating, masking, input/output filtering, and model configuration. Two key features of this layer are a masking module with SAP Data Privacy Integration and the ability to convert modules from an object into an array, creating a fallback chain for model reliability.
A common trap is assigning only management-plane roles, which govern the resource, rather than the required data-plane role. Additionally, when making a call to the endpoint, the URL requires the deployment name rather than the model name, which is a common source of 404 errors.
The SAP model-availability matrix should be checked to ensure the model is available for cross-region load balancing and to ensure EU-only processing if required. Lastly, it is important to note that Cloud Connector is not an egress proxy for arbitrary outbound REST calls to Azure's public endpoint. The complete guide with pricing sources, checklists, and troubleshooting tables can be found on the SAP AI Lab website.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
