As either you’ve heard, noticed, or made use of, Mark Zuckerberg and Facebook have followed the growing communications trend of Silicon Valley and thrown chatbots into Messenger. While at the F8 developer conference, the newly billionized social network opened up their communication product’s API with support for programmatic response entities.
If you don’t use Slack in your office, then it’s probably been a while since you’ve interacted with one. It was probably during AOL Instant Messenger’s height of the mid-2000s that you talked with SmarterChild. It was an extremely rudimentary chat program, but it had the key differentiator to make it a bot: it remembered.
Granted, what it could grasp was basically what you wanted to be called, but its virality (albeit largely based in novelty) proved that people were willing to talk to a digital persona. And if technology has proved anything, ideas often far precede capability. Take a look at the Virtual Boy to the Oculus Rift and HTC Vive, Total Recall‘s Johnny Cab to Google’s self-driving cars, etc.
It’s important to make the distinction between chatbots and things like Siri and Amazon’s Alexa. The method of interaction is key, for sure, as voice and text are vastly different (especially in regards to linguistics), but also they aren’t aiming to interpret your queries and returning with a commensurate response. They take a request and they give a binary reply of either “I did it” or “I can’t do that.”
Those two reasons are why we still operate these sorts of interfaces with at least some layer of abstraction from a completely natural composure. But messenger apps fix this with discrete input. There’s no noise to filter or accents to account for. With text, backend systems can focus purely on NLP, or natural language processing.
Which lands us where we are now, on the precipice of “conversational commerce,” as Developer Experience Lead at Uber Chris Messina calls it. In that Medium piece, he goes into far more detail and analysis than I will about the coming days of chat business, but one of the key takeaways is the idea that personalized, evolving threads with form between you and businesses.
This is vastly different from how we use apps now, which is a singular, normalized interaction with each icon we tap on our homescreen. And no matter how optimized the UX or simplified the UI, you still have to take time with each new app you open to switch mental contexts and reconfigure your desires to fit the way someone else wants to receive your input. That, very obviously, is not ideal.
So you can see why chatbots would be ideal. While the backend would be far more complicated with NLP and layers of AI (artificial intelligence) and ML (machine learning)—the latter of which still has a ways to go as evidenced by Microsoft’s Tay—the user-facing result is far more amenable. You already understand the interface because, ostensibly, you’ve been using it for years: you talk.
Now we arrive at the question of does Messenger and its compatriots in the messenger realm like Slack and Kik fulfill this ideal. Slackbot certainly does. The entire Slack experience is customized through your initial conversation, though it ends there if your domain doesn’t have more bots hooked up.
This is at least partly where Messenger falls apart. It still upholds the archaic structure of siloed intent. You open up a new conversation with each new chatbot, which once again introduces the same brain drain in putting abstraction between your desires and your intent. And worse yet, these early days of the open API has engendered bots that only accept preconfigured missives.
Just go ahead and read through the experience of Trevor McNaughton, cofounder and designer over at init.ai, which specializes in training conversational AIs. Structured messages? Uh, nope! This is supposed to be a conversation, not throwing guesses at an instruction manual. Granted, this will improve as the platform matures, but the key to success actually lies somewhere else.
The optimized version of a chatbot integration into a messaging platform is almost entirely passive, much like how Uber works from within Messenger. It waits until you naturally reach a point in conversations with other people before it starts to work its (admittedly rudimentary) magic. Is that an address, I see? Let’s order an Uber!
These bots needs to work with you, not simply for you. If you mention getting dinner, let the Yelp bot suggest some nearby restaurants that are customized to your past suggestion acceptances. Then, place a reservation with the OpenTable bot, which will then remember some time range of when you like to eat. Both the natural integration and the thread-like memory are vital to making these work.
And then when you just have a blanket request or question, you throw it to a catchall bot like Slackbot. The important thing here is that you don’t have a bunch of different conversations going at once with different objectives of fulfillment. Tell the bot you want to order flowers for a birthday and it’ll route requests to the right place (and it’ll remember this event to suggest again next year). Tell it you want to buy a new toothbrush (and it’ll ask you again in three months if you want another one).
Cofounder and CEO of Kik makes a good argument for singular bots, though, in this VentureBeat op-ed, but the grand sense of futility and dire uselessness in these nascent days of Messenger bots is important to recognize. But it’s just as crucial to understand that this is the direction we’re headed.
Facebook doesn’t create trends anymore. It rides on a wave that they know is going to crest. And after Slack announced its $80 million fund for bot development, Telegram’s rapid platform development, and longstanding bot integrations, Facebook is finally tapping into something a lot of other people saw long ago. But it’s also not sure how to get there.
To which I say: figure it the fuck out. I cannot abide another SmarterChild.