Photo Credit: Shutterstock
For social media marketers, one of the most promising trends of the past 12 months has been the emergence of the chatbot. In a social media world where it’s increasingly difficult for brands to have true 1-on-1 interactions with all of their fans, the chatbot was supposed to be the solution.
The social promise of chatbots
Just think of all the ways that marketers have already experimented with these chatbots. They have built bots to answer questions, make purchase recommendations, help to process orders, send shipping updates and hold engaging conversations with customers. The White House has experimented with bots, and so has Uber.
The big idea at Facebook was that Messenger, its instant messaging platform, would become a platform for ride hailing, mobile payments, door-to-door delivery, gaming and customer service. In 2016, Facebook further opened up Messenger to chatbot developers. One big idea was that, over time, these bots would become as popular as apps: just like iOS or Android has a full ecosystem of apps, Messenger would have an ecosystem of bots.
The bot problem is real
But now Facebook, one of the biggest proponents of the chatbot craze, has said that it’s time to re-think the bot concept. In short, Facebook recently cited a 70% failure rate with its chatbots. That’s a huge number. In a world of Six Sigma, where failure rates of even .0003% are considered intolerable, a failure rate of 70% is basically the equivalent of saying that a bunch of blind monkeys tapping on a keyboard could do a better job of answering customer questions than a bunch of Facebook chatbots.
And it’s not just Facebook that’s reporting problems with chatbots. Microsoft had a mega problem on its hands last year with Tay, its AI-powered chatbot that was supposed to learn from users how to interact on Twitter. But Tay soon turned into a vile, racist, sexist hate-spewing bot. (Hey, it wasn’t Tay’s fault, it was the Internet’s fault!) And the Chinese company Tencent has reported a plague of “scam bots” and “spam bots” on WeChat, its messaging platform.
How to fix the bot problem
The good news is that Facebook plans to take up the bot quality issue at its upcoming F8 Conference in April. The goal is to make it possible to deploy chatbots without requiring a human agent to fill in. Just imagine if you went to a retail store and, for 70% of the questions that you asked, you were directed to the sales associate’s manager. (Where is the shoe section? I don’t know, let me ask my manager. Do you accept credit cards? I don’t know, let me ask my manager). It would get pretty old, pretty fast.
So one idea that Facebook has been batting around is simply giving bots a much narrower set of questions to answer. You can almost think of this like the FAQ of a website – every brand has a set of questions that everyone asks, and that have very defined answers. (What hours are you open today?)
Baby steps for bots
Building on this stable core of questions, it would then be possible to add in more and more functionality. And, as artificial intelligence (AI) becomes more and more powerful, you’d start to see some powerful AI-powered bots, each with a specific brand personality and capable of coming up with original responses, not just a bunch of canned answers.
Instead of tweeting at brands on Twitter, you’d be interacting with bots. Over time, you’d forget that you were interacting with a bunch of computer code, and might even start developing relationships with these bots. That’s when you’d know that the bot problem had finally been fixed.