An Azure search service with built-in artificial intelligence capabilities that enrich information to help identify and explore relevant content at scale.
Yep - you should really consider migrating. Azure OpenAI On Your Data (classic) retired on October 14, 2026, so, as you pointed out, Microsoft no longer supports it. While your underlying models, such as gpt-4o or gpt-4o-mini, may continue to function as raw endpoints, the structural On Your Data orchestration layer, which manages automatic intent generation, data ingestion, vector search, and response synthesis, is deprecated and at risk of breaking.
The risks of continuing to use the retired classic service include a lack of support and stability, since Microsoft may not fix bugs, API degradation, or security issues affecting the legacy retrieval pipeline. New OpenAI models, optimized chunking mechanisms, and faster embedding capabilities may not be made available for the classic architecture. Services that progress from deprecated to fully retired status may eventually have their endpoints shut down.
Microsoft has consolidated its AI tools under the Microsoft Foundry platform. To replace On Your Data, you should migrate your workloads to Foundry Agent Service combined with Foundry IQ.
You will not necessarily have to start from scratch. Microsoft provides a migration framework that offers a number of benefits. The legacy On Your Data architecture is tied to its existing endpoints, whereas the new approach can reuse your existing Azure AI Search index. The new platform offers a broader model catalog, including Azure OpenAI, Mistral, Meta, and other providers. It also provides more flexible orchestration through Foundry Agent Service, with evaluation and debugging capabilities, alongside modern API endpoints and SDK integration.
To proceed, review the Microsoft migration guide, Migrate Azure OpenAI On Your Data to Foundry IQ, to understand how to update your application's retrieval and citation logic. Ensure that you have the required Owner permissions on your resources and a system-assigned managed identity if required by the migration process. You should also plan to transition your codebase from the older azure-ai-inference package to the standardized openai SDK, as appropriate for your target architecture.
If the above response helps answer your question, remember to "Accept Answer" so that others in the community facing similar issues can easily find the solution. Your contribution is highly appreciated.
hth
Marcin