Artificial intelligence and natural language processing (NLP) have already come a long way, yet many users tend to focus on the limitations of these technologies and are quick to judge bots based on the holes in their user experience. Product designers encourage a hybrid of text, buttons and other interactive features to compensate for limitations of NLP and A.I.
However, in every text-based messaging interface, the text input field is the primary mode of interaction. Bots should be prepared to provide a response to any type of input they receive. There are a number of strategies developers can execute to help mitigate dead ends that lead to user frustration and instead optimize their conversational experience.
1. Be up front
Don’t wait for users to input “Hello” or “Hi.” Your bot should lead by setting expectations right away and explaining its capabilities. Keep it brief. Users don’t have the patience to be trained. They already expect that bots should be able to adapt to them. Because of this, there is a good chance users will uncover holes in your conversational experience, but if you can set some expectations right off the bat, you’ll minimize the chance of bottlenecks and create a better path to engagement. Consider explaining how to interact with your bot in one or two short messages and ask the user to type “Help” to learn more about your bot’s capabilities.
2. Get creative with fallback messages
Bot misunderstandings should trigger a fallback message if a direct response cannot be provided. However, generic messages like “Oops I didn’t get that” tend to emphasize the weakness in conversational technologies. Consider multiple fallback messages and rotate them so that any misunderstanding doesn’t come across like a 404 error. If your fallback responses are creative enough, you might just delight users with the unexpected. With Wordhop, we log every misunderstanding and use historical data to tailor our responses to users, but you can easily create an array of clever responses in your code to keep your bot sounding clever even when it fails.
3. Refocus users
If your conversational experience takes an unexpected turn but you’re able to maintain context, you can serve up two messages in quick succession. The first is the fallback message and the second is a reminder of the context. In this sense, you’re able to funnel users back to your goals and not let the misunderstanding become the focus of their experience.
4. Remind users
Remind users of the Help command when your bot fails. It’s an escape path that is universally understood. You can serve up your creative fallback message followed by a reminder that help is just one command away. Remember that showing the Help button too often reminds users that there’s a problem. Inform users at the outset that they can ask for help, and drop in subtle reminders throughout the experience if and when your conversation reaches a dead end.
5. Redirect users
When your bot doesn’t have a response, you can load a fallback message followed by a call-to-action. Ask the user if they would like someone to follow up and, if they agree by confirming “Yes,” forward the user’s message to a CRM solution of your choice and follow up through their connector to your conversational experience. Confirm with the user that the request has been forwarded and that you will get them an answer. It’s hard enough to capture the attention of a user, but rather than lose them to a fail event in your UI, turn the moment into a lead generation opportunity with a follow-up conversation.
6. Tap into external data sources for knowledge
Wikipedia is a very deep and well-structured data source that is also very well maintained. If your bot receives a question it can’t answer, you can tap into data sources like this to make your chatbot seem trained and clever. News sources also provide well-structured, timely data that chatbots can tap into. You can use a headline as a fallback response, followed by a call-to-action asking the user if they want the source of the story. While you may design your bot to have a very limited scope of interaction, these little details are things users will appreciate if they follow down a conversational path that is outside of your bot’s functional specifications.
7. Team up with other bot developers
Build relationships with other bot developers. When your bot doesn’t have a response to provide, maintain a database of other bots that could service the response. Your bot could refer the user to another bot. You’ll provide a value-added service to your users and, as a lead generator for other bots, you could possibly monetize misunderstandings and turn failure into a revenue opportunity. Keep in mind though that this is a strategy that would be prohibited by Facebook Messenger, as cross promotion is against the platform’s policies.
VentureBeatVentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact. Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:
- up-to-date information on the subjects of interest to you
- our newsletters
- gated thought-leader content and discounted access to our prized events, such as Transform
- networking features, and more