Four Types of Bots Used for Customer Chat

customer chat four types of botsWhether you’ve been contacting your cable provider or seeking assistance from an online retailer, chances are you’ve come across a bot used for customer chat. Telltale signs include rapid-fire response time, generic language, and a lack of the pleasantries we’re accustomed to using in human conversation. If you’re lucky, the bot will let you know up front that you’re speaking with a machine. But in most cases, the bot reveal occurs due to a system error, lack of comprehension, emotional understanding, or inability to solve the problem at hand.

From a business and brand perspective, it’s important to know what kinds of bots are being used in your ecosystem – especially if you’re considering an artificial or augmented intelligence solution for your customer chat channel. Whether you’re kicking the tires on a competitor’s technology or just want to know what’s out there, I’ve put together a helpful list of the four bot types that are most often used for customer support and sales force automation.

Menu Based Bot

This bot is easy to spot. Although it exists in the context of a chat or messenger window, it has limited interactivity and functions more like a webpage. The platform enables the type of communication that’s possible, and its capabilities are determined by the entity that owns the platform. After Facebook’s infamous bot fail, the company adopted menu-based bots. Prior, they had bots that were open-ended, allowing users to type an inquiry of any kind in their own natural language. The company quickly found that its AI wasn’t advanced enough to support open-ended text bots, so they switched to menu-based bots that simply let users point and click to match their query.

Tree Based Bot

This type of bot is programmed according to a decision tree, kind of like the dichotomous key you used to identify species in primary school. The logic of this algorithm is pretty straightforward, and there’s no machine learning under the hood. Tree based bots can only contain conditional or control statements (think if-then-else). The difficulty with this type of bot is that there’s no way to know that the language deployed is effective or resonates with the customer. Furthermore, customers may give a whole host of responses that signify ‘true’ (e.g. yes, ya, yup, affirmative, right, correct), making it difficult for the bot to process responses.

The creators of this chatbot are given the impossible task of programming all possible queries and responses for those queries, with a ‘return to home’ option that functions as a safety valve when no response to the query exists. Some tree based bots look for specific words or monitor for sentiment to understand when it’s time to route the customer concern to a human e.g. give me a human, or wtf?!.

Retrieval Based Bot

This is the first bot on our list that actually uses machine learning. Of the machine learning bots, the retrieval based model is a bit more intelligent than its predecessors. The way this bot is built makes it well-suited for automated FAQs. The retrieval based bot holds a bank of possible answers in its memory, which could be based on a list of frequently asked questions. Human coders create a predictive model that pulls up the answer the customer is most likely looking for based on the language they used to make the query. When the retrieval based bot fails, its creators review what went wrong and add more answers to the memory bank.

Generative Bot

Unlike the retrieval based bot, which retrieves answers from a stored bank of phrases, the generative bot actually creates its own language from scratch. These bots are built from models that use deep learning, natural language processing, and natural language generation – all processes that fall under the umbrella of artificial intelligence. They’re typically created from billions of data points, and leverage deep learning because they have to be able to predict the fifth, sixth, and even seventh word in a sentence to generate a clear response. Given the computing power and time it takes to build generative bots, they’re still in relative infancy and can almost never preserve the context in which a conversation occurs. They also have a hard time generating verbs, which have a more dynamic function in human language compared to pronouns, nouns, and adjectives.

Two examples of generative bots include X.AI and Dream. X.AI is a system that can read emails, generate meetings, and automate responses. Despite its apparent complexity, X.AI still requires a human to review the action for clarity. Dream feeds the data generated by the user into a deep learning neural net, which then puts words together in their likeness.

Whether or not you choose to deploy a bot for customer chat depends on the task at hand. For simple tasks like generating calendar appointments or automating an FAQ, chatbots can be useful, effective, automation technology. When it comes to more nuanced operations, such as human conversation, communicating brand voice, and building customer relationships, the most effective approach is to create deep learning algorithms that augment the behavior and emotional intelligence of the humans already doing these jobs.

Contact a RapportBoost.AI team member to learn more about using augmented intelligence for your customer chat channel!

Dr. Michael Housman

About Dr. Michael Housman

Michael has spent his entire career applying state-of-the-art statistical methodologies and econometric techniques to large data-sets in order to drive organizational decision-making and helping companies operate more effectively. Prior to founding RapportBoost.AI, he was the Chief Analytics Officer at Evolv (acquired by Cornerstone OnDemand for $42M in 2015) where he helped architect a machine learning platform capable of mining databases consisting of hundreds of millions of employee records. He was named a 2014 game changer by Workforce magazine for his work. Michael is currently an equity advisor for a half-dozen technology companies based out of the San Francisco bay area: hiQ Labs, Bakround, Interviewed, Performiture, Tenacity, Homebase, and States Title. He was on Tony’s advisory board at Boopsie from 2012 onward. Michael is a noted public speaker and has published his work in a variety of peer-reviewed journals and has had his research profiled by The New York Times, Wall Street Journal, The Economist, and The Atlantic. Dr. Housman received his A.M. and Ph.D. in Applied Economics and Managerial Science from The Wharton School of the University of Pennsylvania and his A.B. from Harvard University.