Chatbot fallback

What a chatbot does when it cannot understand a customer's message or a step fails, such as asking again or passing to a person.

A fallback is the path a chatbot takes when the expected answer does not arrive: the customer types instead of tapping, sends a photo when a number was expected, or your own system does not respond. Every real conversation hits one sooner or later.

Fallbacks decide whether automation feels helpful or frustrating. A bot that repeats 'Sorry, I did not understand' three times loses the customer. A bot that rephrases the question once, then offers a person, keeps them.

Design fallbacks deliberately for each question: what counts as a valid answer, how many times to ask again, and where to go after that. Handle failed API calls with a message the customer can understand and an option to try again or reach a person. In Marichi Talks, Flow Builder conditions and handover steps let you build these paths, and an AI answering step can respond to free questions from your own knowledge base.