The Importance of Modeling User Needs When Creating Chatbots

Image courtesy of rawpixel on Pixabay

2017 saw a huge rise in the adoption of chatbots. Platforms like Facebook Messenger and WeChat drew attention away from traditional social media applications. Each of these messaging clients has a billion users worldwide. Continuing these gains requires satisfying user needs, and being responsive to the limits of this medium. This post highlights a few ideas as you plan for the new year.

Listen to the User

Approximately 45% of end users prefer chatbots as the primary mode of communication for customer service inquires. These users have the same needs as before, they’re just changing the communication tools. In most cases, information already collected and rendered in web and mobile is where the content is. Your challenge is to unlock it. What content is already provided in your FAQ’s? What are common requests your existing customers have? This is an excellent source to get started.

Get started through an introduction, and highlight features.

When starting a new business or product from scratch, traditional user interviews are an excellent place to collect this information. Don’t fall prey to tools as index cards, post-it notes and notebooks suffice. Focus on the user, and what they are trying to accomplish. Interview in person whenever possible to capture context and validate early concepts.

Image courtesy of Engin Akyurt on Pixabay

Leverage Rapid Cycle Development

It’s important to leverage the machine learning deployment model even if your bot is heavily supervised. Frequent releases of code and data into to natural language and event processing enables learning.

Start by deploying a minimal product, then iterate based on what features are being used. It’s important to allocate time to understand what style in language your customers use with a messaging client, including terms and acronyms.

Simplify the User Interface

With text based chatbots, it’s critical to provide shortcuts in the dialog. Typing errors are common on smaller keyboards. Natural language responses are just as important as natural language understanding.

On platforms like Messenger and Slack, use buttons wherever possible. For example, if you’re trying to gather information about an order, try this.

Enable possible choices with one touch.

This reduces the frustration around typing errors, and reduces the transaction time for both you and your customer.

Keep the Conversation Interesting

A chatbot is a very personal experience with a user, so throw away some of the stiff language found in style guides written for a browser. Leverage the natural style found in messaging, including exclamations, and capitalization for a special effect.

Can your bot respond to a “Thumbs-up” response?

Emoticons and images are good, but need to be appropriate with your existing brand guidelines. Also, make sure you are prepared to respond to these references in a dialog. If your bot can’t handle a thumbs up or a smiley face, don’t encourage it by sending these to the user to begin with.

Image courtesy of jill111 on Pixabay

More Details and Learning

If you want to learn more about the underlying technology, check out my latest Chatbot on Facebook Messenger. It uses AWS Lex for NLU processing, and response logic in AWS Lambda.

Here is the code repository on GitHub.

terrenjpeterson/caloriecounter_caloriecounter - AWS Lex based chatbot that calculates calories based on different fast food restaurants. This was an…_github.com