With the 26R1 release, Veeva Safety introduces the Narrative Agent in Vaults using Vault AI for Safety. This generative AI tool refines narrative content by correcting spelling and grammar and improving readability. Depending on your business needs, you can configure your Vault so users can manually run the agent on any narrative document, and automatically when users run the Generate Narrative from Outline action for rules based narratives. The Narrative Agent can:
- Proofread: Update any spelling and grammar issues without changing the structure or details.
- Rewrite: Update the content to improve flow and voice. This may include changes to the structure but should not remove any details.
When the agent runs automatically, both actions execute on the narrative content. When users manually run the agent on an existing narrative document, they can select which actions to run on the content. Narratives generated by the agent may contain up to 100,000 characters, though typically include fewer than 20,000 characters.
You can configure custom instructions and custom actions that Vault applies when the Narrative Agent runs:
- Custom Instructions: Configure the Supplemental Objective field for the Proofread Narrative and Rewrite Narrative agent actions to add custom instructions for a specific action associated with the Narrative Agent. For example, you can configure the Rewrite Narrative action to update chronologies so that the output is ordered by date and update the narrative with a specific company voice.
- Custom Actions: Configure custom agent actions for the Narrative Agent. You can specify when these custom actions automatically run for each Narrative Outline by selecting them in the Narrative Agent Initial Case Options and Narrative Agent Follow-up Case Options fields. For example, you can configure a custom action to summarize the document for medical review in less than 50 words.
Prerequisites
Before enabling this feature in your Vault, you may need to configure the following, depending on your business needs:
Update the Case Layout
To display the agent’s status on the narrative’s associated Case, add the Narrative Generation Status field to the Case layout for the AER, Case, and Imported Case object types in the Narrative section.
Update the Narrative Outline Layout
Add the following fields to the layouts for the Narrative Outline object so that you can define this value on the applicable records.
- Narrative Agent Initial Case Options
- Narrative Agent Follow-up Case Options
Define Narrative Outlines
To automatically run the agent when users run the Generate Narrative from Outline action, update the following fields on the applicable Narrative Outlines. These values specify that the agent automatically runs for Cases associated with the outline for initial Cases, follow-up Cases, or both:
- Narrative Agent Initial Case Options
- Narrative Agent Follow-up Case Options
Configure the Generate Text with Narrative Agent Action
To allow users to manually run the Narrative Agent, configure the Generate Text with Narrative Agent user action on the applicable states of the Narrative Lifecycle document lifecycle. We recommend configuring the action only on the Draft state.
Note: When users run the Generate Text with Narrative Agent action on a narrative generated by narrative templates with merge fields, narratives generated by the Narrative Agent do not maintain any merge field tokens inherited from the template, so they cannot re-render the narrative after running the agent to update the data points. For these narratives, we recommend users wait to run the agent until late in the Case’s lifecycle, after medical review and QA.