Claire

Chatbots have already stepped in as customer service agents for many brands, now they’re coordinating focus groups.

Take Claire, which uses artificial intelligence and chat platforms to conduct product-testing for retailers. The firm was accepted into the latest class of Y Combinator, the Mountain View, Calif.-based firm that funds start-ups.

In its first month at the accelerator, the company has attracted $1.8 million in funding from Salesforce Ventures, Y Combinator and Burch Creative Capital. In August, Claire won Salesforce’s first start-up pitch competition, Dreampitch, with judges such as Will.i.am and Mark Cuban.

The Claire chatbot works on platforms such as Facebook Messenger and Slack to garner input from customers on potential products, finding customers through e-mailed links or through social media and asking a series of open-ended multiple-choice questions. The customer earns points by “playing,” and more points mean more chances of winning something, such as a gift card.

For a recent Rebecca Minkoff product test, for example, the retailer asked the customer’s age, then how many Rebecca Minkoff items the customer bought in the past six months before asking for feedback on five wearable-tech items. The customer was asked, for example, if they would buy the “Always on Power Tassel Keychain” for $50? They could then earn more points for explaining why.

Cofounder Misha Laskin said quizzes have an 80 percent completion rate and customers spend about 14 minutes chatting, with 50 percent coming back for more.

The biggest differentiator between a quiz from Claire and something more traditional, he said, is that “when you ask someone if they’d buy something, that doesn’t mean they will if money is on the line. Focus groups and surveys are notoriously incorrect.”

The machine learning algorithms solve for that by weighing the importance of the answers. For example, a college freshman might say she likes a $300 Rent the Runway dress, but the technology can weigh if that answer is credible.

Laskin, a theoretical physicist, started the company with Marta Jamrozik, whose background is in product development for a consumer packaged goods company. He said that, inspired by the high failure rate for new products, the two were originally thinking about deploying the technology over the web, but then Facebook opened up its Messenger platform to developers, and they witnessed “crazy” engagement rates using chatbots, partly because it felt like a game.

The program combs through the data to determine the demand for each item and makes recommendations on which items are most likely to be top sellers and poor performers. The retailer can view a real-time dashboard.

Claire charges an annual subscription through a tiered system based on the number of products that are tested each quarter. People give opinions on more than 50 products on average per chat session. Eventually, the company is even able to suggest inventory levels for each item.

The company can use Facebook demographic data to target customers who most closely represent the retailer’s demographic, and it can later serve personalized recommendations to the customer based on their answers.

Within the course of a week, the company can deliver feedback, but tests get most responses within the first hour of being sent out.

“We want to enhance designers’ and buyers’ lives by testing more than they would ever be able to do before,” Laskin said. “Before, there were human constraints and manufacturing constraints — but we end up testing the product 10 ways in the same day.”