Chatbots have been hyped as a "revolution" in machine/human interaction, but anyone who has been forced to plod through a bot "conversation" on Facebook Messenger (sadly, I have) knows that the current chatbot reality hardly approaches "revolution" status. Indeed, as Victor Luckerson accurately sneers, "[M]any of them are just clunky repackagings of mobile web pages."
With precious little AI powering today's chatbots, you can be forgiven for thinking chatbots are "overhyped and unimpressive," as one developer survey revealed. This, however, takes the criticism too far. According to an interview with Begin founder and CEO Brian Leroux, "Chatbot hype clearly outstripped the abilities of early bots, but now we're in a quiet productivity period."
After talking with a wide array of chatbot companies, it's clear that chatbots can deliver tangible value now, with the promise of much more to come. The key is to set expectations appropriately.
Despite the overexcitement around chatbots after Facebook's 2016 developer conference, today's best chatbot companies have settled down, focusing on attainable use cases. They do this in a few key ways.
What may be a dirty word in cloud computing fits perfectly with chatbots, but in this case the "hybrid" refers to a blend of human and machine intelligence. The best chatbot companies recognize the current limits of general AI and don't bother. Instead, like Dexter, they might use a bot to take the heavy lifting of information gathering/survey data, and then hand off this data to a live person to follow up (fully informed by that early data collection).
In other words, bots are good where there's a huge range of options that need to be narrowed down, then handed off to a human for the closing experience.
Services like Apple's Siri or Amazon's Alexa deliver general purpose AI. While such services get better all the time, they still fail to understand language that is trivial for humans to grok. As such, smart chatbots like Kylie.ai target narrow knowledge bases, like customer service.
As such, the chatbot need can focus on understanding inputs relevant to, say, a finite product set, thereby ensuring the bot is better able to process those inputs and deliver more meaningful engagement. For example, Init.ai watches interactions between customer service representatives and customers to supply the representative with brand-consistent responses to use, with a focus on ensuring a brand's voice permeates all customer interactions.
Consumers increasingly tune out or turn off app notifications, making it harder for a company to reach them. At the same time, it's hard to escape the app graveyard and earn a spot on a consumer's home screen, particularly given how much time they spend on social and other messaging apps. Rather than fight this trend, however, savvy chatbot companies tend to build on popular platforms like Facebook Messenger. Not only does this give them a potential audience of over one billion daily active users, but consumers won't turn off notifications as readily for their messaging apps, given that they're ground zero for friend and family communication.
SEE: The Complete Machine Learning Bundle (TechRepublic Academy)
While most chatbot companies start with an open-ended command line-style interface, invariably they've moved toward offering multiple choice buttons. As one chatbot exec told me, "Consumers panic if the interaction with the bot is too open-ended."
Talking with Leroux, he stressed the need to iterate on what "easy" means:
Keep simplifying. You think you're building the MVP [Minimum Viable Product] but our customers have shown us the way to the real MVP. They don't want to talk with a robot. They just want to get stuff done. When we started we didn't build something that helped you focus - it kept interrupting to chat.
These aren't the chatbots that were supposed to infiltrate our every digital experience, wisecracking and becoming our best friends as we shop, travel, and more. But precisely because bots have been trimmed to do less, they're now in a position to deliver on (or reset) expectations and become part of a brand's overall customer experience.
- Why an app-focused strategy could lead to mobile failure (TechRepublic)
- Why chatbots like Ask Wiz aren't the future of tech support (TechRepublic)
- AI chatbots are overhyped and unimpressive, say developers (TechRepublic)
- Facebook boosts business appeal of Messenger with new NLP capabilities (TechRepublic)
- Your advertising stinks. Chatbots can help (TechRepublic)
Matt is currently head of the developer ecosystem at Adobe. The views expressed are his own, not those of his employer.
Matt Asay is a veteran technology columnist who has written for CNET, ReadWrite, and other tech media. Asay has also held a variety of executive roles with leading mobile and big data software companies.