Gaining and retaining customers requires a multidisciplinary approach to constantly learning about what they do, what they want, and what aggravates them. That was a key takeaway from a discussion at VentureBeat’s Transform 2018 AI conference among executives at ridesharing service Lyft, community-driven fashion retailer TechStyle, and the latter’s AI provider, Linc Global.
While Linc provides the underlying technology, Aarde Cosseboom, director of GMS technology and workforce management at TechStyle Fashion Group, and his team have to be vigilant to find failures and pain points in customer service. Cosseboom points to breakdowns in the “choose your own adventure” process of working with a customer service bot. “The world changes,” he says, “and maybe A and B can’t satisfy your customer. And maybe there’s an option C that’s growing,” which is discovered when customers are kicked toward a human because the chatbot can’t resolve their problem.
“That’s when we go to our partner and say, ‘Our confidence is low on this intent. Maybe it’s returns‘,” says Cosseboom. TechStyle will take the information to Linc, which does machine learning on the new service path so it can be quickly added and tried out.
Experimentation is essential, explained Monica O’Hara, head of acquisition and growth at Lyft. She is surprised how cautious companies are in deploying AI technologies and encourages them to seek faster insights by trying new tools and approaches.
A key aspect of Lyft’s readiness to innovate, she said, is its cross-departmental collaboration in teams she calls pods. “You explain the problem to a diverse group of people that can solve this problem,” she said. “In our case, that pod consists of a marketer, a data scientist, an engineer, and a product manager.” Those employees sit together every day to encourage natural collaboration across departments.
In the process, employees even cross disciplines. “On my team, we have tremendously technical marketers that in many cases could even pass the bar as a data scientist,” she said. “We have tremendously business-minded engineers who are running my Facebook ad campaigns.”
Customer interfaces also have to break out of silos and be multidisciplinary, said Fang Cheng, CEO of Linc Global. Take the use of Facebook bots, a popular channel for interactions with customers. “When they start to converse with a brand, they are going to go from one topic to another,” said Cheng. They don’t just come to resolve a return issue or to learn more about products, for instance.”
A non-siloed approach facilitates acquiring new customers or saving others from bailing. “We’re using AI not only to support customers in their transactional needs — like ‘Where’s my package’ or ‘Can I return a product?’ But we’re also using [bots] to acquire them again,” says Cosseboom. “Some of our customers [contact us] to cancel, and we actually retain them with our AI machine learning bots.”
And more silo bridging occurs between machine learning on customer interactions and traditional customer relations management (CRM) systems, says Cheng. “When we work with brands, we don’t just work on the problem of what the customer says or what they typed into Facebook,” she explained. “When a customer says they want to exchange shoes for the next size up, integration with CRM can avoid the long back and forth that would take place with a traditional customer-service agent — questions like: “What style of shoes? and “What size did you buy?”
“With the context data from outside of the conversational data … I know you just bought these particular shoes and you bought a particular size and the next size [was] bigger,” said Cheng.
All these efforts to integrate data, disciplines, and learnings from both successes and failures help companies take service beyond basic transactions to something that is proactive and individualized. And better service encourages customers to stay with a brand, even if their faith had been faltering.