“Conversational commerce is a fancy term for what is actually a pretty simple interaction,” says Jenny Fleiss, cofounder and CEO of Jetblack, a personal shopping service. “It is just having a dialogue with someone to make a purchase.” Jetblack, launched in March 2018, uses a mix of text messages, voice-assisted communications, an iOS app, and in-app conversations to process customer orders—a portfolio of touch points Fleiss expects to continue for the near future. “Omnichannel,” she says, referring to the term for a retail strategy that integrates different shopping platforms, “continues to become more and more broad [to include] texts . . . voice, and [used] to compliment a physical location, apps, and an e-commerce website.”
Fleiss’s view is validated by a new study authored by Jetblack that includes a GCIA survey revealing that only about 3% of consumers depend on voice assistants to do the majority of their shopping, though a third of American use them regularly (top use cases include seeking information and streaming music or videos). Jetblack is part of Walmart’s Store No. 8 incubation arm.
Rival Amazon, which developed the Alexa virtual assistant, also seems to take a measured view on the role of voice in e-commerce—for now. Amazon’s top hardware executive, Dave Limp, recently told CNBC that Alexa’s presence in homes was starting to change consumer behaviors that could lead to more shopping but he added, “I don’t think they are necessarily buying more,” he said. Others are more overtly bullish: RBC Capital analyst Mark Mahaney has forecast that Alexa could drive $9.4 billion in incremental annual sales for Amazon by 2021.
The Jetblack paper concludes that voice will become an increasingly important part of the consumers’ shopping habits, but they’ll also text, use shopping apps, go online, and yes, even go into physical stores. The key to voice assistants becoming more useful, ironically, is that consumers need to use them more so that the artificial intelligence software that fuels them will become more adept at understanding natural language patterns. Fleiss, who also cofounded Rent the Runway, spoke exclusively with Fast Company about the results of the study and the future of retail.
Fast Company: The report suggests we are at the very early stages of voice assistants as a tool for transactions.
Jenny Fleiss: We are at maybe in the second inning. Tools and technology have gotten dramatically better around voice technology in the past few years. About 30% of the US population has one of these devices in their homes. What is still very nascent . . . is shopping on those devices. Most consumers are using them to play music and to set timers, interactions that are very robotic in nature. But it can’t [engage] in a layered back-and-forth conversation and for shopping is often it not the most efficient. Consumers are always going to default to what is the most efficient, best way to shop. We should always remember that just because voice or certain tech is new and is the new hot thing, it doesn’t mean it’s actually the best answer.
When we built Jetblack, we chose to do so on text, largely because that’s actually what was most useful to consumers. Consumers can send images [to Jetblack]. You have a back-and-forth dialogue.
FC: What have you learned from texts that can help build a better natural language interface for consumers who may gravitate to using voice assistants for commerce?
JF: There are parts of those interactions that may transition to voice. I think the ones that are most likely would be when you’re replenishing and reordering a product. [Jetblack] knows the SKU of what you’ve purchased before, and so that’s a really quick, efficient way to shop. I think what gets hard in voice is if you want to a recommendation for a gift item and there’s a required back and forth and a series of questions. That’s where the technology is still nascent and developing. I do think we can create that. Machine learning over text is a great way to get started, then in time that can transition over voice, but it has to truly enable those layered back-and-forth conversations—its just going to take time.
FC: It isn’t a secret that one of the major voice assistants is controlled by a major retailer. What are the limitations of that for consumers?
JF: When we launched Jetblack initially, it was just a ton of learning. What do people want? What are people asking for? And very quickly you realize that there’s a very long tail of what people ask for, and unless you want to say no and redirect people, which is a terrible and inefficient experience, then you need to find a way to have a truly universal catalog. We quickly realized that it’s a very hard problem but also an amazing competitive advantage if we can get it right to partner with as many brands and retailers as possible to offer customers the inventory that they want. More than half of the products that people are buying through Jetblack aren’t even available on Amazon, as an example.
We are able to introduce customers to new products and brands. So maybe someone asks for a recommendation for a collapsible car seat, right? We found a brand that’s a much smaller brand, and given that their order volume is still kind of low, they noticed [how] we’ve now ordered a couple of them, and they proactively want to be our partner. It’s a great way to expose and introduce new brands that, if they don’t get placement on a major retailer, might not have made their way into a consumer’s home.
FC: Many consumers still worry that smart speakers are listening to their private conversations. What is Jetblack’s overall view on privacy?
JF: [We’re trying to] build a brand that is trusted, and how that ties into privacy is incredibly important. [Trust] enables you to be as efficient as possible, which enables you to win customer share of wallet. The more information I save about you, likely the better I am able to service you. And so at Jetblack, we think about that as saving your favorite brand, saving that particular SKU of paper towels. But we’re also saving your child’s shoe size, and if you want to reorder a pair of ballet shoes, we can recall which ones those were. But what else do you do to build trust as a business and brand? [We can] educate people about how we take privacy and their information seriously, but there’s also creating other touch points, so that you do have an emotional connection with a consumer, creating a community and a real care and thoughtfulness about knowing you as a person while also respecting that privacy.
FC: How hard is it to train voice assistants to engage in conversations that can lead to recommendations and not just replenishing stock items?
JF: It’s a really big challenge. It’s a hard problem. If you are enabling natural language dialect and you’re enabling a universal catalog, the number of pathways and iterations of purchases is, is so huge, right? It’s, it’s almost infinite. And, and the way that data gets trained is by having much more repeatable use cases so you can learn and train from patterns in behavior. So the broader you go, the more useful it is for the consumer, but also the longer some of the automation will take. Now what we’ve done as a result is we’ve found ways to use technology and automation to make human [customer service] agents faster. So for us right now, everything we do is a combination of agent and automation.