Converting a website visitor into a loyal, high margin customer requires a whole host of conversational skills beyond the simple grammar and semantics. A good live chat agent needs to read between the lines, be able to assess the bigger picture of the sales opportunity at hand, hold multiple parallel conversations at once, and respond to their conversation partner with timing, wit, and emotional intuition. Depending entirely on chatbots as brand ambassadors can be risky.
People do business with people they like, and the best way to get someone to like you is to develop rapport. When a sales person has rapport with a client, they laugh at the same jokes, establish a common ground of experience or expertise, and create a mutual base from which to partner and support each others’ goals. IBM Watson has made some impressive claims and closed some big brands to large enterprise AI sales and marketing contracts, recently drafting thousands of ads for Toyota. However, other chatbot snafus have shown that chatbots are inadequate for most sales applications. Unlike writing copy, customer care requires the dynamic art of two-way conversation. Let’s look at another example: conversational speech recognition. Geoff Zweig at Microsoft research claims that bots have reached parity with humans in recognizing speech with 5.9% accuracy. While this is exciting news, the figure comes with an important caveat. The research parameters instructed bots to transcribe speech from a voice recording rather than take on the full task of conversation, which requires crafting accurate responses on the fly. Two-way conversation is much more challenging to bots.
Chatbots Gone Wrong!
There’s been a significant improvement in chatbot technology, but for e-commerce applications, a live chat agent still reigns superior. From Microsoft’s Tay.ai to Skype’s Translator, a handful of highly publicized – and embarrassing – instances have shown that bots aren’t ready to be the first point of contact with humans. Chatbot snafus are prevalent and there seems to be no end in sight.
Tay.ai started out as any other well-intentioned bot. Designed as a millennial Twitter user targeted toward users between ages 18-24, Tay.ai just wanted to have engaging conversations with other Twitter users her age. Microsoft developed her to ‘get smarter’ as she had an increasing number of conversations. Users engaged Tay on subjects ranging from conspiracy theory to Hitler, and soon enough, Tay made several politically incorrect comments on Twitter before being pulled from the platform.
Earlier this year, Facebook shrunk its AI efforts after its bot API fail rate hit 70%. What does this mean, exactly? In phase one, the Facebook bots were meant to answer users’ questions and insert relevant web links into the conversation. The bot could only manage this task 30% of the time without a human agent. Facebook’s bot was modeled off of Tencent’s WeChat app, which successfully uses chat applications for e-commerce. But what Facebook failed to account for is mobile user behavior – according to VentureBeat 29.1 million Americans still don’t make purchases on their mobile phones, which varies wildly from Asian mobile commerce rates.
It’s not only Facebook that’s scaling back its efforts. Everlane, one of Facebook’s two retail partners, announced that it’s removing several features of its Messenger bot and returning to email Everlane. The one feature that Everlane will keep, is using Facebook messenger as a point of contact between customers and live customer service agents.
Chatbots may be ready to converse without human intervention five plus years from now. But until then, Learn more about the live chat sales training solutions from the team at RapportBoost.AI.
-By Tony Medrano & Meredith Lackey