What an AI receptionist should do when it lacks the answer

A well-built AI receptionist admits when it doesn't know, collects the caller's name and number, and logs the call for staff follow-up. A poorly configured one may guess at prices or hours, and a confident wrong answer can be worse than no answer at all. The article argues that this behavior depends on how the agent is set up rather than on the underlying AI model. Key safeguards include stating the limitation plainly, offering a concrete next step such as a callback, and recording unanswered questions so the business can close the gaps. The author recommends testing agents with simulated callers and checking whether every stated figure appears in the agent's source data.