Sign up for our newsletter and get the latest big data news and analysis.

An Unusual Use For Chatbots – Gathering Data Sets

rob-mayIn this special guest feature, Robert May, CEO and Co-founder of Talla, discusses how companies pile into natural language chatbots, they are overlooking one of the key things a chatbot can do – acquire and grow important data sets. Rob May is the CEO and Co-Founder of Talla, which builds intelligent assistants to help knowledge workers better do their jobs. Previously, Rob was the CEO and Co-Founder of Backupify, (acquired by Datto in 2014). Before that, he held engineering, business development, and management positions at various startups. Rob has a B.S. in Electrical Engineering and a MBA from the University of Kentucky.

In an era of big data, the performance of most artificial intelligence products boils down the size and quality of the data sets they have to work with. So it’s strange to me that, as companies pile into natural language chatbots, they are overlooking one of the key things a chatbot can do – acquire and grow important data sets.

A cursory review of the chatbot landscape will show two primary kinds of bots. The first is a bot that will help you buy something. Since chat interfaces are lower friction interfaces than traditional websites, e-commerce companies are hoping that users will use their phones to message a bot and buy whatever they want. Early indicators though, are that chatbots aren’t that great for most shopping experiences.

The second primary type of bot is one that automates a customer service process. These bots are pre-programmed with FAQs and other tier 1 support materials, and then attempt to answer basic user questions with pre-programmed answers, passing the things they don’t know on to human handlers.

Chat interfaces work well in scenarios where people can type short messages and get quick replies. If that’s true, why can’t we also use chatbots to build out data sets we don’t have, or fill in partially complete data sets? Let’s take the example of filling in customer information on a website. You’ve probably had a debate about how much data to collect on an initial web form. The more data you collect, the better decisions you can make. But the more data a user has to enter, the greater the likelihood that they abandon the form and it isn’t complete.

A chatbot could solve this problem by allowing you to minimize the data collection on an initial form, and then over time, ask for small pieces of data here and there to get a complete data set. This type of lower friction methodology for data collection is easier on the user, and still completes your goal of generating a good complete set of profile data about that user.

Or take the user case of an employee directory in a company. You can see a scenario where you get an email asking you to fill out your employee profile. You click through and see 14 fields, and decide it will take too long so you will do it another time. What if an employee directory bot could pop up once a day via a chat interface and ask you for one simple piece of information? It takes you 10 seconds to do that one small thing, and in less than a month, your profile is done.

With the rise of chat interfaces at work like Slack, Yammer, and HipChat, deploying these tactics in the office is easier than ever. Rather than using bots just to spit out pre-programmed phrases, or perform tasks easily performed in other ways, there is a real opportunity to use bots to tackle the challenge of gathering data sets that were previously difficult to get. The low friction experience of a chat interface is perfect for asking frequent, short, simple questions, and is a highly scalable way to gather data. So when you think about a data set you would love to have from your customers or employees, consider whether a chatbot may be a good way to get it.

 

Sign up for the free insideBIGDATA newsletter.

 

 

 

 

Leave a Comment

*

Resource Links: