Chat

What if I told you there is a bot revolution afoot and and it’s going to change everything we are just now learning about user experience design?

No, I don’t think we are living in the Matrix or Google’s self-driving cars are secretly orchestrating a revolt. What I’m talking about is chatbots. In particular, chatbot development, chatbot UX and chatbot analytics.

Recently, Ilkar Koksal, the co-founder of Botanalytics, spent time with our Growth Marketing class at GrowthX Academy. Botanalytics is a data collection, analysis and visualization tool for conversational interfaces. It records every part of a user’s conversation, displays trends and helps us humans gather actionable insights.

I can already see the comments to this post:  “Chatbots are dumb!” “Chatbots are a fad!”

It kind of makes sense. If I had a dime for every time I cussed out Siri, I would have saved enough to buy a new iPhone 7.

But, here’s the thing: with advancements in natural language processing (like this bot that can predict the onset of psychosis in at-risk youth with 100% accuracy), chatbots are reaching a level of proficiency that allows them to understand what we are saying/typing and respond appropriately. To be fair, a bot has yet to legitimately pass the Turing test but the tech is definitely getting better.

What’s a Chatbot?

A chatbot is a computer program designed to emulate the experience of conversing with a human being.

Generally speaking, there are two types of chatbots:

  1. Rule-based chatbots
  2. Machine learning chatbots

Rule-based chatbots operate based on a set of instructions that are individually predefined. It helps to think of a chatbot’s possible actions as a big decision tree (or workflow). Each tree branch is another possible user-generated answer that must be considered and coded into the bot’s behavior.

For example:

User: I want tea.

Bot (recognizing the trigger “tea”): Do you want “green tea” or “black tea?”

If the user chooses “green tea” then the bot gives one answer, let’s call it “X,” and if the user wants “black tea” then the bot gives a different answer, let’s call it “Y.”

If the user says something other than “green tea” or “black tea,” then the bot defaults to another option (let’s call it “I don’t know wtf you’re talking about, try again.”).

As you can imagine, the process of programming every kind of tea would be tea-dious (excuse the pun). But my point is that rule-based chatbots clearly have strict limitations.

Machine learning chatbots have limitations as well but they are far more capable than their simple, rule-based cousins. These bots use complex algorithms designed to analyze – and learn from – human interactions. Traditionally, the problem is that for these algorithms to work with a high-degree of accuracy, we would need a large enough sample size – thousands of unbiased human conversations of the relevant subject matter.

New technologies, however, are making machine learning algorithm chatbots more effective with a far smaller sample size. What all this means is that we can create better chatbots faster.

Implications for UX

Imagine a world without order forms. A world where your grandparents buy their medications online without your help.  A world where hearing the world ‘hamburger button’ elicits no reactions or thoughts other than hunger. A better world.

Yes, it’s possible. Yes, it’s not too far off.

And you know what’s leading the charge?

You guessed it. Chatbots. If one could talk or type all their requests for products and information through a conversational UI, website navigation would effectively become obsolete. So, would UX designers get kicked to the curb? Not even close.

User experience will still be an in-demand skill except now, instead of designing how a user interacts with visual elements, a UX designer’s job will largely involve designing how a user interacts conversationally.

In the near future, many UX designers will be focused on creating the personalities of chatbots.

Think about what it’s like to talk to any specific person — your mom, your best friend, your favorite store clerk. Now imagine the process of designing an experience that can consistently emulate a personality that elicits a set of feelings and reactions similar to those. Crazy, right? How do even you do that?

Botanalytics

The path to creating lifelike chatbots (just like the path to creating any breakthrough technology) is paved with data.

We have Google Analytics for websites. We have Mixpanel for Apps.

We need data on chatbots to help us figure out what works, what sticks, what actually feels human.

If you can understand when most humans get frustrated conversing chatbots, when they drop off, then you can begin to hypothesize why it is that they drop off. With the right insights, you can consistently iterate by tweaking your chatbot’s responses.

You can essentially be the ghost in the machine that teaches the bot how to act human.

Creating better bots will result in partial automations of the sales cycle, reducing the time salespeople spend on repetitive tasks. Bots will also enable faster troubleshooting of technological problems, alleviating much of the burden placed on tech support specialists. And we all know what that means: better experience, more conversions, improved customer satisfaction and higher (and more predictable) revenue.