AI Deep Research: Workflow-Initiated Assessments

Important:

This feature is available for customers with solution packaging.

Overview

The AI embedded flow for workflow-initiated assessments helps you accelerate third-party assessments by generating a deep research report on the specific third party being evaluated within an existing workflow. By using publicly available data to source relevant answers, the feature reduces manual research effort, helps teams move through assessments more efficiently, and strengthens the value of workflow automation in the assessment process.

This use case is best suited for customers who want AI embedded in existing assessment automation. Deep research runs automatically when a workflow stage rule launches an assessment, making this ideal for high-volume, repeatable processes like standardized onboarding flows, vendor stage transitions, or reassessments.

Prerequisites

  • Auto-responses for assessments enabled under AI features in Global Settings. For more information, see Managing AI Features in the OneTrust Platform.

  • Vendor assessment template with AI auto-response option enabled for the AI experience. Templates set to manual are not eligible.

  • Ensure the vendor record has a domain or website. The system uses domain first and website as a fallback.

Configuring the vendor template

  1. On the Third-Party Risk Management menu, select Setup > Templates. The Vendor templates appears.

  2. Select your desired assessment template. The Template details screen appears.

    Important:

    The template must be newly created (unpublished) or a new version of an existing template.

  3. Go to the Details tab.

  4. In the AI experience section, select the AI auto-populate option under Auto-responses.

    Tip:

    Enable Scan past assessments for relevant information to scan the three most recent assessments connected to the same primary record and use relevant past responses as context for AI suggestions or auto-population.



  5. Save and publish the template.

Configuring the workflow stage rule

  1. On the Third-Party Risk Management menu, select Setup > Workflows. The Workflows appears.

  2. Click the View button on the Vendor workflows tile.

  3. Select your desired workflow. The Vendor workflow details screen appears.

  4. Select on the desired workflow stage where you want the stage rule to trigger.

  5. In the Stage Rules tab, create a new rule that either:

    • launches a new assessment and select the assessment template that you previously configured for AI-driven responses.

    • copies a previous assessment and select the assessment template that you previously configured for AI-driven responses.

      Note:

      If a previous completed assessment exists and Copy Latest Assessment's Responses option is selected, the copied responses take precedence and AI response automation is not applied.

      If no previous completed assessment exists, or if responses are not copied, deep research can still run to populate the new assessment.

  6. Select the AI response automation setting.



  7. Save the stage rule configuration.

  8. Publish the workflow.

  9. When a vendor workflow advances to this stage, the stage rule will trigger and launch the assessment.

  10. After the assessment is created, OneTrust checks the vendor record for a domain first and uses the website only if no domain is available.

    • If one of those values is present, the system calls the AI response automation flow, generates a deep research document for the vendor, and uses that document to auto-populate questionnaire responses. This runs asynchronously, so the stage rule and assessment launch do not wait for AI processing to finish.

      The end result is a launched assessment with AI-populated answers already filled in for review, and the deep research report is attached to the assessment as supporting evidence.

    • If neither a domain nor website is available, OneTrust still launches the assessment, but deep research and AI auto-population are skipped for that run. In that case, the end result is a standard launched assessment with no AI-generated answers.