How clarification questions work
An assistant should clarify when the answer or action would change depending on the interpretation: two locations, two services with similar names, or a garbled number on a phone line. It should not clarify what is already clear, such as asking a customer to reconfirm a time they just chose.
Good clarifications are one short question with the likely options, asked once. On calls, they matter even more, because speech recognition can mishear names, numbers, and email addresses, so spelling back critical details is a form of clarification.
Example: “I have no hot water”
A caller tells a plumbing company, “I have no hot water.” The fix depends on the heater type, so the assistant asks one question: “Is your water heater gas or electric?” The caller says gas and mentions a pilot light, so the assistant reads the company’s relighting safety note and books a visit for a technician. This scenario is illustrative.
Clarifying in InsertChat
InsertChat’s response settings include “Ask a clarifying question when the request is unclear,” on by default. Turn it off and the assistant asks only what the next approved action requires. Either way, it clarifies only missing or ambiguous details and doesn’t ask customers to reconfirm a clear request, chosen time, or name.
On calls, the AI receptionist asks one short clarification when words are unclear, never guesses values from unclear audio, and spells back email addresses for confirmation. See the plumbing setup.
