Retail’s chatbots are evolving, and bringing new challenges

The word “chatbot” holds something of a…negative connotation for many retail shoppers today. Interacting with a bot on a website is notoriously frustrating, slow, and unhelpful.

But the world of chatbots is changing, and fast. The scripted bots of just a few years ago are out, and there’s a new sheriff robot in town.

Today’s chatbots are software systems that use a branch of AI called Natural Language Processing to understand and respond to everyday human language. (If your mind went straight to ChatGPT, you’re on the right track!)

The widespread adoption of conversational AI could bring efficiency and improved customer experience to the retail world, addressing everything from supply-chain woes to onboarding issues. But despite the large number of AI offerings out there, the rapid evolution of retail chatbots hasn’t come without challenges.

Changing landscape

The advancement of chatbots has happened faster than people realize, Raghu Ravinutala, CEO and co-founder of conversational AI platform Yellow.ai, told Retail Brew. It was just a few years ago that rules-based bots (which were limited to pre-programmed responses and often used for FAQs) were popular, he explained.

AI chatbots, on the other hand, rely on language models (tools that analyze language and predict the most useful response), and can generate human-like responses. AI chatbots can also be trained and improved to offer better results over time.

“[Today’s AI chatbots] are very advanced,” Ravintulata explained, pointing to customer service as the top retail use case for conversational AI. “They can read sentences, they can manage context. We’ve seen companies that process 60%–70% of their customer interactions [via chatbot] with high [customer satisfaction] scores.”

Broader impact: Customer support isn’t the only area in which retailers are experimenting with new and improved chatbots.

Take Walmart, for example. In December, the retail giant launched “text to shop,” a chatbot built in-house that allows shoppers to search for items and checkout via text message.

Separately, the company is automating supplier procurement negotiations with the help of Pactum AI, whose chatbot negotiates with human suppliers on behalf of companies.

In a 2021 pilot conducted in Canada, Walmart asked the bot to negotiate payment schedules with partners who supplied products used, but not sold, in stores (like carts). The chatbot reached a successful agreement with 64% of suppliers, which was well above the 20% it said would yield a positive ROI.

Pactum’s chatbot can simultaneously conduct thousands of deals, addressing contracts that are usually left by the wayside, Pactum CEO and co-founder Martin Rand told Retail Brew.

Logistics, labor, and commodity prices increased during the pandemic, but as prices stabilized, retailers with over 100,000 suppliers often didn’t have the capacity to manually re-negotiate each contract to adjust prices, Rand explained.

“If we were able to normalize this fluctuation in prices, we are essentially becoming a counter-inflationary force in this situation,” he said.

One to teach them all: In 2020, San Francisco-based research laboratory OpenAI released an API (application programming interface) making it possible for apps to be built using its language processing model, GPT-3.

• GPT-3 is one of the largest and most powerful language processing models out there today, trained on 300 billion words from around the internet. It’s also the model behind viral sensation ChatGPT.

When it comes to the evolution of chatbots, there’s the world before GPT-3, and the world after GPT-3, explained Vasant Dhar, a professor at the NYU Stern Business School.

“Fundamentally, [chatbots have] changed, because for the first time, you have the ability to talk to a machine on your terms, as opposed to its terms,” Dhar said.

Essentially, GPT-3 has made it easier for retailers to build virtual assistants, doing everything from making recommendations and checking inventory to order tracking, and setting up curbside pickup.

Not so fast: The possibilities for conversational AI in the world of retail are immense, but retailers have several challenges to consider.

For one thing, consumer behavior might not be ready for the new era of chatbots.

“It’s going to be really interesting to watch the industry figure out how to get people to re-engage with [chatbots],” said Justin Keller, VP of revenue marketing at conversational marketing platform Drift. But prior failed chatbot implementations are weighing against them, he added.

“The pressure’s really on now—you’re not gonna be able to get away with having a bad experience twice,” Keller said.

Dhar warns that companies should be wary of jumping on a trend for the sake of it.

“There’s a need to fundamentally rethink, ‘Where am I getting advantage?’” Dhar said. “Your competitive advantage is not customer service; everyone has that,” he added.

“It’s not enough to say, ‘Well, I’ll adopt GPT-3 and the world will look hunky-dory,’ because everyone else will adopt GPT-3 too. That just becomes kind of part of being in business.”