Fallback Detection

Most chatbots embarrass you by answering anyway. Ask something outside what they know, and they produce a confident, wrong answer - the exact behaviour that makes teams afraid to put AI in front of customers. Wonderchat is built to do the opposite: recognise the edge of what it knows and stop there.

When a question falls outside what your content actually covers, the agent detects it and says so, then hands off to a human or falls back to the path you set - rather than inventing an answer to fill the silence. Because every answer is grounded in your own content and cited to its source, the agent knows what it's standing on; when there's nothing underneath a question, it doesn't pretend there is.

This is the single feature that decides whether you can trust an agent in public. An assistant that guesses is a liability. One that knows its own limits and escalates cleanly is something you can actually deploy.

You control what happens at the boundary - hand off to live chat, capture the question for review, or return a fallback message. Every out-of-scope question also becomes a signal about where your content has gaps. It's part of a SOC 2 Type II certified, GDPR-compliant platform. Put an agent in front of customers that knows when to say, let me get someone.