Failure, in short, is a certainty in life. There’s always a chance that things will break down, that people will make mistakes. And chatbots are no exception. They can break, misunderstand, and make errors.
But what’s important is how those failures are dealt with. A chatbot failure, though not ideal, doesn’t always need to lead to a bad user experience.
So, here’s a guide to making even your chatbot failures as user friendly as possible.
Chatbot failure can cover a wide range of incidents in which a customer service bot fails to meet the mark.
This means it can be anything from completely breaking down, to tonal faux pas.
For instance, the chatbot may fail technically. This would cover issues such as broken connections to needed databases, issues with the server or volume of requests, and so on.
Chatbot failure can also cover things like broken record messages – when a chatbot fails to understand and repeats the same error messages over and over. It also includes tonal mistakes, like the finance chatbot which used ‘rapey’ language.
Other examples are chatbots that give nonsensical replies, or questionable suggestions.
In short, chatbot failure refers to any time that a chatbot’s communication breaks down — for any reason.
The key to boosting the UX of any potential issue is understanding the impact that it could have on the customer/user.
So, what’s the impact of chatbot failure on its users?
In truth, you need look no further than the angry chatbot feedback found on Twitter to see the impact of a poorly handled chatbot failure. Frustration abounds, along with increasingly heated requests for human support.
On top of that, chatbot failure often means that if users manage to get in touch another way (and they should be able to do so easily) they may well end up repeating themselves. No one likes repeating themselves.
It’s clear, then, that chatbot failure can damage the user experience. But it’s also not possible to guarantee that there will never be a technical hiccup that results in some form of error. How, then, can you improve the user experience of such a frustrating situation?
1. Know what can go wrong
You can’t seek to prevent things — or fix them quickly — if you don’t know what could go wrong. So, mark every possible point of failure that may befall your chatbot by testing it against real user queries.
By knowing what could go wrong, you can plan to minimise the effect it has on the user as much as possible.
2. Have good chatbot error messages
Don’t keep users in the dark, and don’t deny there’s a problem. Both are frustrating. Acknowledge the chatbot failure.
The point of an error message is to let the user know there has been an error — and to get the conversation going again.
Make sure your error messages offer a helpful explanation to the user. They need to clarify the problem in an easily accessible way. That doesn’t mean a cold, mechanical ‘server error’ notice and accompanying error code.
Rather, make sure the bot has a message to fall back on when encountering any of the points of failure you’ve identified. Highlight where the error comes from so that the user can avoid triggering it again.
3. Always offer a solution/workaround
It’s imperative that users aren’t trapped with a failing chatbot. This means supplying options in times of chatbot failure.
For example, the bot could suggest other topics the user can talk about with them. (By clarifying its abilities, or even offering buttons for topics to get things moving.)
You should also offer an escape route. Make it easy for users to abandon the failing bot and connect with a human to chat.
4. Turn off that which is broken
If the chatbot failure stems from a major issue that renders it unusable and in need of a major fix, turn it off. There’s no point in offering a chatbot that can’t help anyone — all that will do is frustrate more people.
5. Monitor your chatbot
If your chatbot were to fail, right now, would you know? A chatbot should not go mute, and the key to managing chatbot failure is minimising the impact and time it takes to solve the issue.
All this is to say, monitor your chatbot. Set up alerts for errors that might impact your bot. Analyse its performance regularly. In short, keep an eye on what your bot is up to, so you can intervene before a chatbot failure becomes a service failure.
6. Don’t be complacent
Collect feedback from your users, and use it. If there’s a common issue, address it to make your chatbot more robust.
Keep learning and improving. Tune your chatbot: is the tone of voice correct? Is there an error you missed that needs an error message drafted?
Whether it’s the chatbot misunderstanding, or a technical fault that’s caused your chatbot to fail, it’s not a good experience for your users.
So, how do you mitigate a bot fail? Simply, treat it as you would a human misunderstanding, by apologising, seeking to clarify, and offering alternate backup options – then learning from the experience wherever possible.