Sensitive questions
Some questions carry more weight than others. "Is this guaranteed?", "Can I get a refund after 30 days?", "What does the bereavement policy cover?". A confident but wrong answer here is not a small mistake. Once a company's assistant has said it, customers reasonably treat it as a commitment.
ChatHQ assistants know where to stop.
How the assistant behaves
- It answers only what your documents say — On guarantees, refunds, pricing, legal terms, HR policies and anything involving money, the assistant sticks to the wording in your sources. It does not paraphrase a policy into something broader than it is.
- It does not fill gaps — If your refund policy does not mention a situation, the assistant says the policy does not cover it rather than inferring an answer.
- It points to the official source or a person — For anything with consequences, the reply includes where to read the full policy or who to contact, so the visitor is never left relying on a chat message alone.
What you can do
- Upload the official text — Put the actual policy, terms or price list in the knowledge base. The assistant can only be as precise as its sources. See Knowledge base.
- Name the escalation path — In your instructions, say who handles what: "For refund decisions, ask the visitor to email billing@example.com." See Instructions & behavior.
- Keep policies current — Remove superseded versions. If two versions of a policy exist in the knowledge base, the assistant may cite the old one. See Source citations.
The assistant is not a substitute for your policy page or your team. Treat it as a guide to your official sources, and make sure those sources are the ones you would stand behind.
Test it
Ask the hard questions yourself in the test chat: "Is this guaranteed?", "Can I cancel after the deadline?", "Will you cover the cost if it breaks?". Check that each answer quotes your policy, admits what the policy does not say, and names where to go next.
Related
Was this article helpful?
