Rob - Fotolia
Meet your friendly service agent -- The sales chatbot
AI chatbots enable companies to automate some sales and service tasks, but the combination of people and smart machines creates better outcomes than either one alone.
Read part one of this article, "AI chatbot apps to infiltrate businesses sooner than you think."
Artificial intelligence-based sales chatbots are positioned to replace the live sales assistants responsible for highly repeatable tasks, the same way word processors replaced typing pools. Companies need to upskill those types of employees or plan for an upheaval.
Though cognitive computing tools like the sales chatbot automate some job responsibilities, the technology won't be a wholesale replacement for the sales team because it isn't able to handle novel, complex tasks, according to technology industry analysts, and even the companies that deliver bot platforms.
Forrester Research recently predicted cognitive technologies, such as robots, artificial intelligence, machine learning and automation, will lead to 16% of U.S. jobs being replaced by 2025, while the equivalent of 9% of jobs will be created -- a net loss of just 7%. Office and administrative support staff will be disrupted the earliest, while new roles, such as robot monitoring professionals, data scientists, automation specialists and content curators, will make up some of the 8.9 million new jobs in the U.S.
If you consider how AI chatbots work, you understand why they won't displace more jobs: machine learning platforms only work on repeatable tasks that a person trains them to handle. A sales chatbot can't handle novel scenarios that require out-of-the-box thinking.
That said, people and smart machines are better together. Deep neural networks can identify patterns a person probably wouldn't notice because they occur infrequently. However, the technology is far from perfect, and the machines may provide probabilistic conclusions that require quality control checks, Gartner VP Tom Austin said during a recent webinar.
"There's a human-machine symbiosis where humans and machines complementing one another can outperform either alone," Austin said during the webinar. "Burn that into your head and never forget it; the biggest value that can come out of this is [the understanding that] machines are making humans smarter and humans are making machines smarter."
Indeed, even IBM is careful not to position its crown jewel, Watson, as a replacement for actual employees. Watson augments human intelligence, working side-by-side with humans to enhance their ability to act with confidence and authority, according to IBM Watson Platform Director Steve Abrams.
"What if you could just talk to the data?" Abrams said. "So, in the same way you would ask your assistant for the status of something, your assistant could ask the virtual assistant for the info -- and no one is put out of work."
IBM's Watson Conversation is being put to use in AI chatbot applications that supplement certain customer service and query tasks, and the latest iteration can decipher customer tones and intentions. This lets contact center agents off the hook for very basic questions that are asked again and again, freeing them up to focus on more complex problems.
"We aren't just looking at the literal intention of speech, but also the tone, sentiment and emotion analysis," Abrams said. "I can create a customer service solution that can tell when a customer is becoming irate and tailor the answers, using an apologetic tone, and hand it off to a human agent."
Sales chatbots do the grunt work
Conversica LLC's virtual assistant, a combination of artificial intelligence technologies, is delivered as software as a service. The virtual assistant creates an email dialog to gather information from inbound leads, which is ultimately passed on to live marketing and sales reps, explained Gary Gerber, Conversica's head of marketing.
One of Conversica's customers is KnowledgeVision Systems Inc., an online business presentation platform provider based in Lincoln, Mass. KnowledgeVision's sales chatbot, which the company named Caitlin Kelly, plays an important role in the sales process by following up on low-priority leads that the company's business development representatives (BDRs) don't have time to focus on, according to Susan Zaney, the company's VP of marketing.
"Our BDRs were cherry-picking leads based on which ones they thought would be the easiest to convert to a sale," Zaney said. "We had interesting people giving us an email address, but they weren't getting followed up on because it was tough to tell if they were real or not. This eliminates that problem because Caitlin follows up on every lead."
KnowledgeVision's leads sit in Salesforce, but sales reps don't see those names. First, leads are scored based on whether the company and title are real and how much research the sales team member did. High-priority leads are followed up on by a BDR, but it is a big jump from being a priority lead to actually wanting to talk to a sales rep, Zaney explained.
"Caitlin has bridged that gap," Zaney said. "With Caitlin, we are converting 8% of our untouched leads, and we are re-engaging accounts that hadn't closed, but had potential."
Caitlin doesn't get to the bottom of why someone is interested in a product, though; her goal is only to get potential customers who are hovering at the top of the sales funnel to request a meeting. When someone agrees, she will confirm their phone number and ensure a sales rep has the basic details needed to start a live chat, Zaney said. And if Caitlin learns through the messaging process that someone isn't interested in a meeting or doesn't want more information, that's just as important because it means BDRs don't have to spend time making those calls, Zaney said.
However, the sales chatbot isn't perfect.
"Sometimes she doesn't understand out-of-office responses," Zaney said. "For example, if your out-of-office response says you will return on Tuesday of next week, she may interpret that as you want a meeting on Tuesday."
That means KnowledgeVision's team has to review Caitlin's interactions before following up with customers. But the upsides far outweigh the limitations, according to Zaney.
One major advantage is that virtual assistants are 24/7 employees. If a potential customer responds to an email at 1 a.m., they can respond back immediately, and customers may never be the wiser that this persistent, tireless sales person is actually an AI chatbot.
Sales chatbots serve up fast answers
While these types of conversational interfaces certainly help companies increase customer engagement, they aren't mature enough to take over as the de facto user interaction platform, IBM's Abrams said.
But that's to be expected at this point in the natural language processing evolution. New user interaction platforms have come about throughout history, and during those transitions, the old ways of doing things remained useful, Abrams said.
"Today, I wouldn't feel comfortable booking travel through a conversation platform, and I don't with an actual travel agent, either, because there are so many options that are better to understand visually [on travel websites]," Abrams said. "The limited bandwidth of a dialog channel is better for addressing something like a travel disruption."
If Watson Conversation determines a customer needs to find information from a large pool of data, it can hand the request off to a system that is better-suited for that task. For instance, if a doctor requests a research paper, Watson Conversation can hand the query off to an adjacent Retrieve and Rank service on IBM's Bluemix developer cloud.
"In the medical example, [Watson] is not replacing anyone; it is allowing doctors to get answers faster," Abrams said.
Welltok Inc., an IBM partner, uses Watson machine learning and cognitive computing technologies to drive its CaféWell Concierge platform, which provides healthcare consumers with the information they need. The concierge understands customer intent and provides answers or directs customers to the resources they may need, said Jeff Cohen, the company's co-founder and vice president of functional architecture.
Cohen pointed out that you need subject matter experts to train machine learning systems. At Welltok, employees who have spent time answering customer calls help to train Watson, repurposing their expertise to scale through machine learning systems.
"Now the information isn't sitting in one person's head, and the system is a single access point for all customers to access," Cohen said.
