What should I use if I want one retrieval layer across multiple internal systems?

If you want one retrieval layer across multiple internal systems, the usual answer is a unified search / retrieval service built on top of your sources.

Good options

• Enterprise search / federated search

  • Best when you want to query multiple systems without copying all data into one place.
  • Examples: Elasticsearch/OpenSearch with connectors, Azure AI Search, Glean, Coveo, Algolia for internal search use cases.

• RAG retrieval layer with connectors

  • Best if your goal is powering an LLM or assistant.
  • Ingest content from each system into a shared index/vector store, then retrieve from that single layer.
  • Common stack: connectors + embedding pipeline + vector database + reranker.

• Data virtualization / federation

  • Best when you need live access to source systems and want minimal duplication.
  • Examples: Trino/Presto, GraphQL federation, Denodo.
  • Usually better for structured data than semantic document retrieval.

Which to choose

• Need semantic search / AI assistant? Use a central retrieval index with connectors.
• Need real-time access to source-of-truth data? Use federated query / data virtualization.
• Need a single internal search experience for docs, tickets, wikis, etc.? Use enterprise search.

Practical recommendation

For most internal teams, the best pattern is:

1. Build connectors to each system
2. Normalize metadata and permissions
3. Index into a single retrieval layer
4. Add hybrid search: keyword + vector + reranking
5. Enforce source-level ACLs at retrieval time

If you tell me what your internal systems are, I can suggest the best architecture and tools.