The rise of chatbots has been recent and substantial. B2B and B2C sites now have bots at the ready, asking how they can help you find information or do what you need to do — and they are often on a first-name basis with you.
Chatbots have proven to be one of the ideal use cases for artificial intelligence (AI) and the appeal is obvious. Unlike their human counterparts, chatbots are available 24/7, can simultaneously answer hundreds of queries instantly and come with a one-time cost compared to the continuous cost of employing customer service representatives. Although there is fear that chatbots aren’t as accurate as humans, those concerns have been mitigated in recent years as chatbots can now resolve 80% of customer queries without human supervision. And when they get a question they don’t know the answer to, the chatbot is able to move the customer to a live representative for resolution.
Chatbots are without a doubt one of the most hyped technologies going right now. It seems everyone has either deployed one, or is in the process of doing so. But building a chatbot is not a project that should be approached lightly. There are three key questions you should answer before you start the process.
What Purpose Will Your Chatbot Serve?
Regardless of your industry, one of the most challenging parts of building a chatbot — and where a lot of failures occur — is in deciding what the chatbot will do.
When emerging technology is involved, identifying a good use case can be challenging. Often enough, the lure of the emerging technology is strong enough that organizations don’t actually think about the need or use and instead just dive in. FOMO is real, pushing people to ignore the proper processes of defining and developing. This is counterproductive as users react negatively to the product. This negative reaction in turn impacts the confidence in (and often budget for) further development and FOMO can all too easily turn into a “this technology doesn’t work” attitude and abandonment of said emerging technology.
A lack of understanding of how new technology impacts user interaction works is common. As a result, assumptions are made.
For instance, when mobile apps were first introduced, a lot of developers attempted to put the content and functionality of their entire website into a mobile app. Needless to say, that wasn't successful. The same is now happening with some chatbot deployments. If a mobile app is suited to only take on a subset of web content, a chatbot should be even more restricted and should be viewed as something to take on functionality that may currently be outside of the website (e.g. customer service).
A good use case is one where the chatbot is easy to use and does what it needs to do. Its goal can be to provide information that is otherwise difficult to find, for instance, pointing the user to relevant information in an information-rich site or an information-rich environment, such as flight reservations. Or the goal is to execute a repetitive task such as reviewing structured legal documents or scheduling a meeting.
Related Article: A Good Chatbot Is Hard to Find
What Kind of Chatbot Should You Build?
Chatbots come in many different varieties and identifying the right approach is easier said than done. Once you’ve defined what your use case is, however, things become more clear. The key to deciding on the type of chatbot you need to build is to understand what you want to communicate, how you want to communicate it and to whom you will be communicating.
Is your chatbot communicating with internal employees (e.g. meeting scheduler, HR activities) or with external B2B or B2C parties? The platform you use will be dictated by your audience. There are a host of platform options. For example, if you are using Slack internally, a Slackbot may be a great option. Facebook Messenger continues to be a popular B2C option and B2B chatbots could be embedded in a website.
How will the user interact with your chatbot — will they use their voice to speak or will they type out requests? The decision depends on where the user is located and the kind of query. For instance, a noisy manufacturing floor is not the best option for voice-initiated requests. If your bot covers a broad set of topics that are complex in nature, a menu-driven bot, where users select a predefined option might be more successful than a natural language interface that allows the user to phrase their question any way they wish.
Once you have defined the use case, and based on the use case the audience, platform and method of interaction, you are well on your way.
Related Article: Lessons Learned From a Chatbot Failure
What Is Your Chatbot’s Persona and Tone?
Yes, your chatbot should have its very own persona. Giving your chatbot a name is part of this persona. This will help personify the chatbot and help users identify with it. Think of Bank of America’s Erica, for example. It helps to provide a human element to an AI entity, and can also be an additional source of branding (i.e. repeating the last part of "America" in the name "Erica").
Tone of voice and personality are also integral parts of chatbot creation. These factors not only help define your chatbot, but have the added benefit of distinguishing your chatbot from competitive versions. Tone is extremely important and can make or break the entire experience. The sassy tone of financial chatbot Cleo, for example, is a major part of its identity. It launched the self-dubbed "savage mode" in celebration of Valentine’s Day — as a way of instituting some financial tough love to users. However, the mode wasn’t without controversy at its launch, showing you should be careful with the personality you choose. Yes, chatbots are meant to be more informal, but they should still accurately represent your brand and the users you are trying to appeal to.
The potential of chatbots, along with other AI-infused applications, is very real. Though the hype is deserved, you shouldn’t jump straight into building a chatbot without first answering the above questions. This will ensure you are meeting real customer needs and increasing the chances your chatbot will be deployed and received successfully.