Ever since humans first puzzled over the tangle of vacuum tubes, diodes, and relays which comprised the ENIAC, we’ve found computing technology difficult to use. So difficult, in fact, we invented a paid profession to deal with it: we called these people programmers. Today, we think of programmers as people who make software for users–but at the dawn of computing, programmers were the users. Technology was that difficult.
Fast forward seven decades. The most obvious advancement we’ve made is to put these house-sized machines in our pockets. Less obviously, but equally as important, we programmers have made computing accessible to folks not quite as geeky as us. We did this simplifying human-computer interaction to a shared language of metaphors: the iconography, mental models, and lexicons of the graphical user interface.
Although that shared language was an enormous advancement, it still required users to learn something before they got started. The promise of AI-powered chat and voice is experiences so natural they require little to not learning at all: if you can talk to a human, you can talk to Alexa.
At least, that’s the potential. Many of the chat & voice experiences being designed right now deliver on exactly none of that potential–and worse–they erect even higher barriers to technology. Let me give you an example.
Suppose a restaurant has retained you to build their chatbot. The worst (and most natural) thing to do might simply be to repurpose their existing content: convert their menu into a card gallery or series of dense message bubbles. But we have to ask ourselves–is this any better than just showing a list?
Humans have been looking at long lists of stuff for a really long time. We’ve come up with clever ways to them accessible, many of which show up in ecommerce best practices. The upshot is to balance breadth with depth, so that a user understands the categories of things you offer, along with concrete examples.
But consider the chatbot example I’ve shown here. Instead of presenting a scannable menu, the user is forced to digest the menu in chunks: only after they specify they’re interested in a pre-built pizza, a custom pizza, or sandwiches, are they presented with the details. Why is this terrible? Because users don’t always know what they want–in fact, they rarely do. To determine whether I’d like a custom pizza, I’m probably going to want to see how tasty your pre-built pizzas are. And before I settle on a pre-built pizza, I’d like to compare it to the options for a custom pizza. To complete that simple mental calculation, a pop-up list requires scrolling at most. The chatbot, on the other hand, forces users to click or type at least three times.
So just show the damn menu!
How can we do better?
Remember the fundamental promise of chatbot technology is to make interactions more human, and therefore easier. So a rule of thumb is to ask what a human would do. When you walk into a restaurant, does the waiter ask whether you’d like chicken, steak, or fish tonight? Of course not. They ask if they can help make a personalized recommendation or present you with specials. They provide an interaction layer above the menu itself.
What if the restaurant’s chatbot said something like this:
- “Hi, I’ll be your server today. You can see the menu by clicking below, or I can tell you about our amazing specials today. [Buttons: Menu, Specials]”.
- Or, even better: “Hi there–I see you’re a big fan of Mexican food. Would you like to hear about our award-winning enchiladas? [Buttons: Sure, tell me more, See menu]”.
That experience starts to deliver on potential of chatbots. It does not cram a web page into text bubbles or card galleries. It presents a fresh ordering experience built from the ground up on the promise of chat AI.
To sum up, the first way to Make Chatbots Terrible:
1. Make the user experience worse.
Go forth and build better 🙂 In my next piece, I’ll examine the second of three ways to make terrible chatbots: promise too much. Subscribe to our bot below by clicking “Send to Messenger” to be notified when the next article comes out.