If the popularity of Siri, Alexa and Google Now is any indication, chatbots are here to stay. As noted in Part 1of this series, we believe successful chatbots will keep the design simple and “conversation” to a minimum.
We also believe the most successful chatbots will be utilitarian in nature, and specialize in specific tasks, provide recommendations and excel at helping users complete those tasks more rapidly. And we predict that chatbots will evolve from only performing rote and repetitive tasks, to learning over time and offering personalized interactions with recommendations. This will be possible through their ability to access data, process it and respond quickly, using technologies such as neural networks and machine learning.
To accomplish this, chatbots will increasingly rely on self-learning algorithms and graph databases to help them comprehend larger representations of data without ambiguity. To speed and streamline retrieval, it is important to dynamically update these semantic graphs, which store data types and their relationships. This is key to improving response quality.
Key Design Elements
To begin designing for the bot age, businesses need to peek into the future and design chatbots to work in today’s context. Success will require the following design elements:
Reduced manual effort. The reduction of manual effort required for human interaction is arguably the key element of future chatbot design. This would essentially mean completely removing or at least reducing the number of touches, keystrokes or mouse clicks required to help the bot determine the best solution to the problem. One way to achieve this is to make sure most available options are provided by the chatbot itself, with the user just needing to select the right option. This would greatly reduce the time and effort to interact with the chatbot.
Ability to predict the right options. For chatbots to display clickable options for users to choose from, the right options need to be predicted from a set of choices. This requires the system to understand the user’s problem. The goal is to decipher the problem with the least number of questions from users, which requires an emphasis on understanding the user’s background and the context of his or her request.
Personalization. This leads to one of the most important design decisions: How much personalization can be brought into the interaction itself? For instance, can the system remember user profiles, previous interactions, the interactions of other users in the system, and the current context of the problem? Each of these attributes must be understood in conjunction with the other to really understand users’ needs.
Ability to determine the best approach for unresolved queries. How will the chatbot respond when it doesn’t understand the user’s question or the user doesn’t provide the expected response? Because repeatedly asking users to rephrase a question or try again would be a sure path to frustration, the design here is especially tricky. The chatbot needs to demonstrate a partial understanding to reveal which part of the question it doesn’t understand. For example, “I understand you might be interested in xxx, and I want to help you, but what do you mean by yyy?”
User recommendations. If users are unsure of what they are asking for — or if the conversation enters an endless loop — the chatbot should provide recommendations, such as what Google Allo does when helping first-time users.
Entering the chatbot market is fairly easy to accomplish today. Businesses can quickly design and deploy a rudimentary chatbot and develop a following. That said, many of today’s chatbots have failed to live up to expectations, mostly due to slow technology advancements.
How well does the bot know the user? How well does it know what the user needs? Can the bot of the future be a master bot that assists, advises and continuously learns about the user? Can it be intelligent enough to become wiser but never cross the line by pretending to be human? The future of chatbots, after all, will hinge on their ability to become useful, maybe even indispensable, to human beings by supporting customers and making their lives easier.