Yes, this has been the gradual evolution of AI context and tooling. Same thing is occurring with some of the use cases of a vector DB and RAG. Once you can have the agent interact with the already existing conventional data store using existing queries, there is no point in introducing that work flow for inference.