12 February 2026

It is now possible to provide alternative names for a document type which can be used to classify the document. This is useful when the document type name is an abbreviation or not descriptive. Alternative names also increase the chances of the model correctly classifying the document as the desired type.

You can now create default chained tags for images which is useful for running prompts on image extracts systematically when the image type is detected.

You can now create on-demand agents which can be run from Vault. On-demand also offer more flexibility as you don't need to pre-define the scope of the analysis.

Once you have created your on-demand agent, specify the instruction you would like it to run.

Then within the Vault, select the documents you'd like to run the agent on and click on Run agent.

You'll then be able to select which agent you'd like to use and see the processing in the sidebar.

The results are then accessible from the Agents section or by clicking on the completed run in Vault. The results from consecutive runs will append in the on-demand agent's results summary.

Extracted tables can be included in Agent reports. This allows the reports to be leverage more data than the tags. You can now also extract tables manually from image files.

You can now import a zip or folders into Vault. Vault will recreate the folder structure based on what was uploaded.

You can email attachments to Vault directly. Each Vault has its own email and emails can be refreshed or disabled on demand. You can choose to include the email body in the analysis. When enabled, Vault will be able to answer questions about the contents of the email (e.g. instructions).

The Ariba integration allows you to connect your Ariba documents to Vault and make the reverse export easier with intelligent Ariba supplier matching.

When the Supplier Data API is enabled, Vault will match supplier names identified in documents to their corresponding Ariba IDs.