Verify Responses use cases
This guide provides real-world examples of how to use the Verify Responses feature to build trust, meet compliance requirements, and improve your AI agent.
Verify Responses is especially valuable for businesses in regulated industries like finance, healthcare, legal, where accuracy and audit trails are essential. But any organization that wants to ensure AI reliability can benefit from these use cases.
Get stakeholder sign-off
Before deploying an AI agent, your legal, compliance, or security teams may need proof that it works correctly. Verify Responses gives you the documentation they need.
How to use it: Run sample queries through your agent, click the shield icon, and share the Claim Verifier and Trust Score results with stakeholders. The six-stakeholder analysis (including Legal Compliance, Security IT, and Risk Compliance) addresses their concerns directly.
Example: A healthcare company tests their patient FAQ agent with questions about medication dosages. They share Verify responses results showing all claims are verified against approved medical documentation, and the Trust Score is Approved. The compliance team signs off on deployment.
Build audit trails
Regulated industries often require documentation of how AI systems make decisions. Verify Responses creates a record of every claim and its source.
How to use it: Periodically verify responses from your live agent. Save or export the results to maintain an audit trail showing that AI answers are grounded in your approved source documents.
Example: A financial services firm uses an AI agent to answer questions about investment products. They run Verify responses on a sample of queries each week and store the results, creating documentation for regulatory audits.
Test before launch
Before making your agent public, use Verify Responses to catch potential issues — unsupported claims, compliance risks, or gaps in your knowledge base.
How to use it: Create a list of common questions your agent will receive. Run each one and verify the response. Look for Non-verified claims or Flagged/Blocked Trust Scores, and address them before launch.
Example: A law firm tests their client intake agent with 50 typical questions. Verify responses flags three claims that aren't supported by their uploaded documents. They add the missing source materials and re-test until all claims are verified.
Identify knowledge base gaps
When claims show as Non-verified, it often means your knowledge base is missing important information. Verify Responses helps you find and fill these gaps.
How to use it: Review Non-verified claims regularly. For each one, decide whether to add the missing source document to your knowledge base, or update your agent's persona to avoid making unsupported claims.
Example: An e-commerce company notices their product support agent makes claims about warranty terms that aren't verified. They upload their warranty policy document, and future responses about warranties become verified.
Validate sensitive responses
For high-stakes queries like legal advice, medical information, financial guidance, you may want to verify individual responses before sharing them with users.
How to use it: When a user asks a sensitive question, run Verify responses before providing the answer externally. Check that all claims are verified and the Trust Score is Approved.
Example: A legal tech company uses an AI agent to help users understand contract terms. For complex questions, the support team verifies the response first. If the Trust Score shows Flagged from the Legal Compliance stakeholder, they add appropriate disclaimers before sharing.
Train your team
Verify Responses helps team members understand how the AI works and builds confidence in using it.
How to use it: During onboarding, show new team members how to use Verify responses. Walk through examples of Approved, Flagged, and Blocked results so they understand when AI answers can be trusted and when human review is needed.
Example: A customer success team uses Verify responses during training to show new hires exactly where the AI gets its information. This builds confidence and helps them know when to escalate questions to senior staff.
Related articles
Updated about 2 hours ago