Spring chatbot

SAN FRANCISCO — The robots have arrived, but they’re not ready for prime-time.

“Conversational commerce,” where bots sell brands on chat-based platforms, is getting much of the buzz in Silicon Valley these days, draining away some of the excitement from the still under-realized social media buy button. But experts said chatbots are still better at answering customer service queries than racking up sales.

That could change, as the bots and anything using artificial intelligence get serious attention from the tippy top of the tech world.

Mark Zuckerberg opened up Facebook’s Messenger platform to chatbot developers in April, demonstrating the potential of chatbots on the app, which has a billion users, by ordering flowers. The interaction mimicked standard customer service chat, but there was a robot on the 1-800-Flowers end of the conversation.

Since then, developers have built an estimated 18,000 chatbots for Messenger. They join the legions of chatbots and artificial intelligence-powered tools interacting with users on platforms such as Amazon, Apple, Google, Microsoft, WeChat and Slack.

Tommy Hilfiger introduced a chatbot on Messenger to promote the Tommy x Gigi collection in September. Burberry’s Messenger chatbot took users “behind the scenes” of its recent collection and let users shop three items from the runway show. (It also lets users chat with a person.) E-commerce app Spring in May introduced an experimental feature that guided users, helping them discover brands through Messenger while combining both automated responses and the ability to connect directly with a Spring “concierge.”

And PayPal recently started to slowly link with Messenger to add PayPal as a payment option within Messenger, which means that merchants that use PayPal can accept payments directly in their bots.

Bots use scripting software to automate tasks, such as answering basic customer service questions and searching through product listings. A brand’s chatbot on Messenger can, for example, incorporate features such as images and multiple choice questions and it can also use artificial intelligence, which makes them “smarter” over time based on a user’s behavior.

When flummoxed, they can punt the conversation to a person.

Conversational commerce, though, is still in its early stages and is benefiting from at least some of the online dreams that have fallen flat along with buy buttons, which allows social media users to buy right from their Twitter, Facebook or Instagram feeds, and have largely failed to live up to expectations.

With more than 1.7 billion active users, Facebook is growing into more of a commercial driver given the sheer size of its audience. Much of that comes not from buy button directly on the site, but referrals or ads that tout products and point users to an e-commerce site or a brand’s app.

Forrester analyst Jessica Liu said the excitement around buy buttons have sagged some because people generally don’t turn to social media to make purchases.

“Social networks and brands have a long hill to climb before users become accustomed to transacting on social media,” Liu said. “They’ll need to train users to consider social networks as a point of purchase by making the transition from discover to explore to buy stages seamless and consistent.”

So can chatbots pick up where buy buttons left off? Not just yet. Most say it’s too soon to expect much more than experimental, limited offerings.

Spring, for example, is tweaking its initial approach. “We’re currently working to build an even better experience,” said Spring cofounder and chief executive officer Alan Tisch. “We strongly believe in the platform, and are very excited to launch our improved bot soon.”

Burberry’s bot redirects users to the brand’s mobile web site if they want to actually buy something.

Everlane was one of the first fashion brands to try corresponding with customers on Messenger. The e-tailer lets customers opt-in to receive automated order and shipping confirmations and tracking updates, while the company’s human customer experience team uses Messenger to help customers with orders and answer product questions.

Jon Epstein, chief marketing officer at artificial intelligence software company Sentient Technologies, said there is still a lot of work to be done.

“Chatbots face the same ‘uncanny valley’ problem that 3-D animation did — who can forget the creepy Tom Hanks in ‘Polar Express?’ It’s almost, but not quite, real, which is unsettling. Many chatbots don’t really use AI, but are more rule-based, like voice response systems.”

He expects companies will overcome this problem as more and better AI comes into play. So far, chatbots have proven better at customer service and shipping than actually selling, he said.

The bot shop on the Kik messaging platform has a Victoria’s Secret Pink bot that helps users “find your perfect bra by chatting” while the Sephora bot will teach users about makeup and help them find products used in tutorials and the H&M bot will build an outfit around an item via chat.

Omar Pera, cofounder of Reply.ai, a start-up that helps businesses build and manage chatbots, argued that bots are better than calling a customer service hotline because they’re more convenient and because phone calls are no longer as common.

Pera and cofounder Clara de Soto said that bots on Messenger can help customer satisfaction jump from 10 percent to 90 percent. They advise against bots that pretend to be human.

“People like to trick the user, but a bot should act like a bot,” Pera said.

He added that ultimately, the bot could absorb today’s buy button, meaning, for example, that if a Facebook user clicks on a product ad on his or his News Feed, the user will be redirected to a chatbot that reveals more details and incorporates the “buy” option within the chat.

But for now, de Soto added, stick with the low-hanging fruit of customer support. “There’s a lot of commerce potential, but you can’t just dive right in — that’s more like, the seventh date,” she said.