(WxOR6) Multi-tenant agents with user namespace isolation with deployment rings

Consider automation service providers delivering well known business processes to multiple clients, ex: “order to cash” (O2C). The service provider want to develop one standardised O2C digital agent, A, to deliver the O2C service to multiple enterprises, namely, E_1, E_2, …, E_k. When the digital agent A receives a request r from a client E_i, it needs to use the set of policy documents corresponding to the enterprise E_i and and take actions on invoices belonging to E_i. In other words, the agent acts like a multi-tenant agent with different namespaces corresponding to E_1, E_2, …, E_k. Further, each namespace E_i corresponds to the set of policy documents of E_i and the actual dataset (of invoices) of E_i.

The requirement is for Watson Orchestrate to support deployment of such multi-tenant agents with namespace separation of enterprises using the agent. This essentially means that each namespace corresponds to the data and the set of policy documents of corresponding enterprise. To serve a request, the data and RAG pipeline of the process would run on data and documents corresponding to the namespace of requesting enterprise.

The service providers would also want the agent to support deployment rings by which the clients can be on different versions of the agents based on their policy to switch to new versions.