f t i +

Chatbots, AI, and User Experience

Michael Peach

Research on chatbot UX is starting to reach a critical mass. At the same time, chatbot technology is now affordable enough for small and mid-size companies. Not surprisingly, a lot of companies now want to sell you a chatbot very badly, regardless of whether it’ll improve your customer experience or not.


solutions_conversational-marketingWe’re not here to say you shouldn’t be excited, but honestly - there are limits to what chatbot technology can do right now. Chatbots can be immensely valuable, so long as you play to their strengths and acknowledge their limitations.

To cut through the vast sea of marketing nonsense on this topic available on the internet, we have to start by unpacking the difference between a chatbot and AI.

My Chatbot Runs on AI!

AI is a lot of things: a pretty exciting concept, a movie with Haley Joel Osment, so on. Sadly, it’s not a magic bullet that will fix your website. It’s also not what you’re getting when you buy a plug and play chatbot.

Examples of AI-driven intelligent assistants include Alexa, Google Assistant, and Siri. They take input from the user, and use a natural language processor to interpret the inquiry. Using the same natural language capabilities, intelligent assistants generate a unique response to the question. The technology involved in real deal AI is highly advanced and, famously, still not known for creating great user experiences.

Your standard chatbot isn’t driven by AI. The ones that claim to be usually don’t have the capability to process information and create realistic natural language. Chatbots, generally, are a whole different thing.

Well, How do Chatbots Work Then?

To understand how chatbots work, you need to understand branching logic. If you read a Choose Your Own Adventure book as a kid, you’ve got a head start. If not, here’s how they work:

You, the reader, are the main character.

Rather than being a passive observer, you make a series of choices that drive narrative progression. Do you cross the river or ride the hot air balloon?  

Cross the river and you set off a chain of events that culminates in one outcome — you’re eaten by a tiger, for example. Take a hot air balloon instead and your story changes.

Bot Gif

Chatbots operate on a similar premise (fewer tigers though). They offer the user a series of choices, and their responses determine what comes next. The chatbot has a preset response ready for a handful of possible choices. The problem is that in order to seem like an AI that can respond to any input, many chatbots hide what those choices are.

The outcome? “I’m sorry, I didn’t quite get that”

Super satisfying, right?

But never fear - there are a lot of things chatbots are amazing for, as long as you’re upfront with users about what to expect.  

The best bots guide customers through predetermined workflows to create a desired outcome. They aren’t designed to take unique information and come up with unique solutions, and generally the good ones will escalate situations they can’t understand to a human user as quickly as possible.

Chatbots aren’t intended to be intelligent. A well designed chatbot guides users through a simple linear flow. Just like a Choose Your Own Adventure book, the available outcomes are set ahead of time.  

In order to properly use chatbots, you need to keep this in mind.

Three Simple Rules For Dating My Teenage Chatbot

  • Put UX first. The first rule of chatbots is also the first rule of websites! Always think about the interactions you are designing from the user’s perspective. What questions will create a pathway from the first interaction to the final goal? What kind of experiences are going to be wildly frustrating? How would you want to be treated by a chatbot? As a good rule of thumb, your users will probably have roughly half the patience you have for your chatbot - keep that in mind.

UX Planning

  • Don’t pretend your bot is anything but a bot. Don’t try to fool people when you’re trying to help them. Transparency is key to healthy relationships, and also, your bot isn’t going to fool anybody. People will have different expectations for bots than they do for people. They will also interact differently (read: more efficiently) with this expectation in place.

  • Start small, start smart. Plan for capacity growth, but don’t overpromise. Overly ambitious chatbot projects usually lead to frustrated customers. Unless you want to call your chatbot “experimental” or beta feature, keep it simple. Build concrete, multiple choice pathways that set clear expectations. Make your chatbot do something useful and within its wheelhouse, like open a support ticket or schedule a meeting based on a preset calendar.

Chatbots are incredibly powerful marketing, customer support, and lead qualification tools. But before you deploy: understand how they work, know where AI comes into play, and follow best practices for positive ROI and enhanced user experience. 



Learn more about chatbot logic and conversational marketing strategies from a certified expert, HubSpot's Brian Bagdasarian. 

Check out the interview with Brian + US CEO Mike Skeehan

Definitely not spam

Sign up for our newsletter

Don't worry - we only average, like, two emojis per subject line.

Got a question for Michael Peach?

Message the author of this post and they'll get back to you.

Fire Away