A private AI workspace over your own approved documents

An internal assistant that answers from documents you approved, cites its sources, and respects the permissions your team already has.

What a private workspace is

It is an assistant with a defined library: the handbooks, contracts, specifications and procedures your organisation has approved. Staff ask ordinary questions and get an answer with a link to the passage it came from, so the answer can be checked before it is used.

The value is not novelty. It is that a new colleague can find the current version of a policy without interrupting someone, and that the answer points at the document rather than replacing it.

Architecture options in plain language

There are three practical shapes. A hosted workspace in a tenancy dedicated to you is fastest to stand up. A private-cloud deployment in your own account keeps documents and logs under your contracts. A self-hosted deployment inside your network suits data that may not leave the premises at all, and costs more to operate.

The choice follows your data rules and your operational capacity, not the model. Whichever shape you pick, retrieval works the same way: documents are indexed, a question retrieves the relevant passages, and the answer is generated from those passages with the citation attached.

Access control and privacy questions worth asking

  • Which documents are in the library, and who decides when one is retired?
  • Does the assistant inherit each person's existing permissions, or is the library shared?
  • Where are questions and answers logged, for how long, and who can read them?
  • Is any content used to train a model? For our deployments, the answer is no by default.
  • What happens when a document changes: is the index rebuilt, and how quickly?

Pilot checklist

  • Pick one department and one document set with a clear owner
  • Write 20 real questions staff actually ask, with the answers they expect
  • Run those questions before launch and read the citations, not just the answers
  • Decide the escalation path for a wrong or missing answer
  • Review after two weeks of real use: what was asked, what failed, what to add

Built on WorkenAI

Our private AI workspace deployments run on WorkenAI: any model, your own documents (RAG), budgets, guardrails and full audit trail, in the cloud, hybrid or on-premise.

What we will show you

  • A demonstration built on a document set you supply, with citations visible in every answer
  • The 20-question review sheet, filled in, including the questions the assistant answered badly
  • A written data-handling summary for the deployment shape you choose

Questions

Will our documents be used to train a model?

No. Documents are indexed for retrieval and stay in the deployment you choose; training on your content is not part of the service.

Can it respect our existing permissions?

Yes, where your systems expose them. Permission inheritance is one of the first things to confirm during scoping, because it changes the architecture.

How do we know an answer is correct?

Every answer carries the passage it came from. The pilot review reads citations rather than trusting the summary, and that is also how you find gaps in the library.

Request a scoped pilot

Tell us the workflow, who approves the output and which systems hold the data. We reply with a scope, the open questions and what a first pilot would need from your side.

Related