Everyone has experienced that moment in a conversation when it’s not clear how they should reply. For chatbots, these moments result in chatbot error messages.
Chatbot error messages are a potential chink in your chatbot armour. Unless they’re done well, they can ruin an otherwise great chatbot experience.
So, how do you get them right?
What are chatbot error messages?
Also known as fallback responses or default responses, chatbot error messages are what your chatbot says when it cannot respond to a user’s message.
This may be due to conversational divergences (where users deliberately try to test the boundaries of your bot), typos, or simply an incompatible request. Whatever the reason, your chatbot doesn’t know how to reply to the user’s question. But it still needs to say something.
That’s where chatbot error messages come in.
Guide away from the error
First things first, the point of an error message is to let the user know there’s been an error. More, it’s there to get the conversation back on track.
This means the user needs to know what caused the issue. Otherwise, they have no way of knowing how to proceed with the conversation, and so they’ll bounce. The worst chatbot error messages are ones that ask the user to ‘try again’, with no indication of how to get the answer they need. If the user does so, they’ll only get another fallback response.
It’s the chatbot’s job to guide the user out of the problem.
An effective chatbot error message will highlight where the error comes from. (Without, that is, overloading the user with too much detail.) You can do this by being open and honest, and clarifying the chatbot’s abilities.
For example, if your chatbot is great at assisting with account issues but can’t handle job applications, then have it say so.
Keep it friendly
Tone of voice is a major component in effective chatbot error messages. To paraphrase a cliché, it’s not just what your chatbot error messages say, it’s how your chatbot says it. Get it wrong, and a chatbot can go from being friendly and helpful, to robotic and disinterested. And that’s infuriating to your customers.
Hitting an error message means something has gone wrong, but your chatbot doesn’t need to be negative about it. The worst default messages imply the user is at fault — a passive or accusatory tone can inadvertently place blame.
For instance, “I didn’t quite catch that” implies the user has done something wrong, whereas a message containing ‘Sorry!’ can remove any hint of blame from the user. So, keep your chatbot chirpy. Let it apologise for not understanding, and be co-operative by explicitly prompting a response that will garner a successful reply from the bot.
Use call-to-actions
A call to action is a clear, simple way to help chatters get back on the conversational track with a chatbot. They tell users what to do or say to get the chatbot chatting helpfully again.
Your call to actions could take the form of buttons with possible topics the chatbot can discuss. Or the chatbot could suggest an input for the customer to use. (For example, ‘help’ to access more advice on talking to the chatbot.)
Have variations
There’s a chance that your user could end up triggering your chatbot error messages more than once. But hitting the never-ending chatbot ‘I don’t understand loop’ is infuriating. When every input results in the same error message, it makes the bot seem incompetent and destroys any sense of a human-like tone.
It’s important then, that there’s some variation to them.
Just as you don’t want your customers to repeat themselves, it’s not great for your chatbot to do so either.
So, make sure to give your chatbot a few different fallback responses to choose from. You could even tailor your chatbots error messages based on whether the customer has had a fallback response previously in the conversation. For instance, offering an escalation to a human after three error messages.
Give users an out
Even great chatbot error messages can get grating if you keep running into them. Sometimes, customers will have a query or conversational need that your chatbot just can’t answer. Hitting that dead-end means leaving the conversation — and the issue unresolved.
But it doesn’t have to. A final best practice tip is to ensure that your customers have an effective out.
This could be to have your chatbot offer an escalation option. Or it could offer to provide other contact information, so the user knows where they can go to find the answers or support that they need.
Chatbot error messages
Good chatbot error messages optimise the chance of getting the conversation back on track. That means a helpful, positive tone, with a clear next step for the customer to take.
Useful links
- - The importance of chatbot tone of voice
- - Have you been suckered by the chatbot vs live chat debate?
- - Why your standalone chatbot is failing
- - A roundup of the worst chatbot feedback on Twitter, and what to learn from it
- - The pros and cons of full chatbot disclosure
- - The problem with generalist chatbots