Live shopping, where potential buyers watch live product demos and ask questions on the spot, is rocketing in popularity even as the principal sales format shifts from television to online platforms. Viewers congregate to see items up close and buy immediately if they like what they see.
Yet consumers often have questions before they buy. In the existing format, broadcasters or streamers—those facilitating the sales—are limited by time and capacity constraints. They can only field a finite number of interactions and inquiries, yet consumers expect immediate answers to their questions.
An artificial intelligence (AI) streaming assistant developed by Nina Huang, a business technology professor at the University of Miami Patti and Allan Herbert Business School, and her research colleagues, aims to mitigate this tension in online selling.
“The AI streaming assistant can provide tailored answers to each customer at the same time, complementing the streamer or influencer to provide a better, more informed shopping experience,” explained Huang, whose research focuses on how digital technology can enhance user experiences and improve business outcomes.
The team’s recently published research suggests that access to an AI streaming assistant increases consumers’ perception that they are receiving pertinent information and reduces uncertainty in decision-making.
A randomized field experiment conducted on Alibaba—one of the world’s largest e-commerce marketplaces—revealed that implementing an AI streaming assistant increased sales by 3 percent and reduced product return rates by 12.55 percent.
Huang was part of the wave of researchers who began exploring the business application of AI technology five years ago.
“I had an interaction with AI-powered customer service and was blown away with how much AI can improve consumer experiences,” she said. “I was betting on the future of AI and started to get my hands on possible business applications of AI technology in different scenarios.
“Live streaming as a type of business model is big and growing in the U.S. market, so I was super excited with this high-stake experiment to have some real impact on consumer purchase.”
Huang noted that the research, which contributes to the literature on human-AI interactions, explored two distinct modes of that exchange: proactive and reactive. In the proactive mode, a consumer watching the stream asks a question in the chat, and the AI assistant responds; in the reactive mode, the AI assistant initiates the interaction by querying the consumer: “You just asked a question, do you want me to respond?”
Correlational results show that these interaction modes reinforce each other in increasing purchases and reducing product return rates, she noted.
The AI assistants are “trained” using scripts and algorithms from prior Q&A streaming sessions.
“It’s not perfect, but the whole point is to provide individualized, timely, tailored responses, using natural language processing technology in AI. When a customer asks you a question, [the AI assistant] provides an answer based on my training data,” she explained.
As an indication of the growing trend to live online shopping, Huang referenced the report that QVC TV laid off 900 employees to pivot their TV shopping to an online live shopping business model.
Huang’s research collaborators include Lingli Wang from Renmin University of China; Yumei He from Tulane University; De Liu from the University of Minnesota; Xunhua Guo from Tsinghua University; Yan Sun from the Alibaba Group; and Guoqing Chen from Tsinghua University.
As external consultants, Huang and her team derive no financial gain from their research.
“We get to know the answers and satisfy our curiosity—this is the reward of scientific inquiry and the spirit of public research,” she said.
She’s especially proud that the findings indicate a downturn in returns.
“That’s one of the biggest headaches for online sellers,” she noted. “When people are not happy, you have to deal with the returns—and that’s a huge cost, in terms of transportation and impact on the environment as well.
“I want to see other websites trying to offer this type of service so that consumers can make better decisions with lower returns,” Huang added. “I’d love to see more use or application of this type of technology.”
