Every case you run collects a set of real documents — bank statements, agreements, policies, ID proofs, and more. Training a case turns those documents into a private knowledge base, so you can ask questions about that case’s files in plain language instead of opening and reading each one.
For example, after training a loan case you could ask “What’s the applicant’s monthly income?” or “Which address is on the Aadhaar?” and get an answer drawn straight from the uploaded documents.
Train a case
Open any case from Pending Review, All Cases, or a flow’s case list. In the action bar at the top of the case, click Train.
Click Train
The Train button sits in the top-right action bar of the case. (If the case has been trained before, the button reads Retrain.)
Confirm
A dialog explains that Xenora will prepare the case’s documents for AI chat, and that this may take a few minutes. Click Yes, train to start.
Wait for training to finish
The button shows Training… while it works. You don’t have to wait on the page — you can carry on with other work and come back later. Training usually finishes in a few minutes; large cases with many documents take a little longer.
Only documents that have actually been uploaded to the case are trained. If a client uploads more files later, use Retrain to bring the knowledge base up to date.
Asking questions about a trained case
Once training finishes, an Ask AI button appears on the case. Click it to jump straight into a chat scoped to that case’s documents.
You can also reach every trained case from the Document Sources page in the AI section of the sidebar — open the Cases Source tab to see them listed alongside their document count. From the Chat page, trained cases appear under the Cases tab of the knowledge-base selector.
Keeping a trained case current
A trained case is a snapshot of the documents at the moment you trained it. When the documents change — a new upload, a re-uploaded correction, an added requirement — click Retrain on the case to refresh the knowledge base so answers reflect the latest files.
Train a case once all its key documents are in. You’ll get the most useful answers from a complete case, and you avoid retraining repeatedly as files trickle in.
Train a Case vs. a Document Source
Both give you a knowledge base to chat with — the difference is where the documents come from:
