While you and your team might have worked hard to create a fantastic conversational experience with great conversation design, seamless integrations and human handovers at all the right places, there will eventually come a time when a user asks a question that your chatbot is unable to answer.
Therefore, it is important for you to be able to provide the user with the best fallback response that can guide them to how the chatbot can help and also help ensure that you’re able to learn from these failures.
Acknowledge the failure
First things first, when your chatbot doesn’t have a clear answer to a question, acknowledge the shortcoming and let the user know by saying something like ‘Sorry, I didn’t quite understand that.’
Guide the user in the right direction
The purpose of a fallback response is to inform the user that something went wrong, but also try to get the conversation back on course. The worst chatbot fallback responses are those that instruct the user to ‘try again’, with no clue about what they need to amend in order to obtain the information they require. If simply ‘trying again’ gets a similar fallback answer, it is likely that the user will give up.
The best fallback responses will point out the source of the problem, as well as outlining the chatbot’s capabilities. For instance, if your chatbot is designed to handle account setup rather than purchases, then let the user know this.
Be friendly
The chatbot’s tone of voice is always the key to communicating with your audience effectively, and the same is true with your fallback response. When a chatbot responds with a fallback message, it usually signifies that something has gone wrong, but this doesn’t have to be a negative experience.
With your fallback response you want to eliminate any suggestion that the user is to blame and to keep it friendly, for instance ‘Sorry, I’m still learning and didn’t quite catch that’.
Point the user in the right direction
Providing buttons or quick replies are straightforward, easy-to-understand approaches that can help users to get back on track with a chatbot.
For example, your buttons might list potentially relevant subjects that the chatbot can help with. Alternatively, the chatbot might recommend inputs that users can try, for instance, ‘Type “help” to get additional information about how to best communicate with our chatbot.’
Provide variations
It’s possible that a user will trigger your chatbot fallback response many times. If every input generates an identical response, the chatbot will appear inept and any feeling that this is like dealing with a human will be lost.
Try to ensure that your chatbot can provide a variety of fallback responses. The best practice is to customise your chatbot’s fallback response based on the context of the conversation or on what responses the user has previously received. For example, the third fallback response could offer to escalate the user’s query to a human, which brings us onto the final piece of advice…
Provide the user with a way out
Even the best chatbots cannot address every possible user requirement and there are scenarios where providing repeating fallback responses is not going to help. There will come a point when it is crucial to offer the user options to escalate their query.
These can include:
- handing the user over to a human via a live-chat program
- collecting the user’s contact information and raising a ticket for a response
- providing details or contact information to direct the user to a channel for a more detailed enquiry.
