Canadian-born fashion retailer Frank + Oak has added an AI-enabled recommendation tool to its online offering, hoping to deepen its credentials as a leading edge e-commerce player.

The brand debuted as an online-only men’s fashion retailer that quickly earned a reputation for its strong customer service and advice on what to wear – no small feat for a business that sold clothes without fitting rooms. It has since expanded into women’s apparel and brick-and-mortar retail, but its new AI service is another online anchor for the brand.

The site has long made product recommendations to visitors. But as part of its new Style Plan subscription service (which recommends clothes that shoppers can piece together into ensembles for $89 per month), the new AI uses purchase history and customer input to offer what Frank + Oak co-founder Ethan Song hopes will be better choices for each individual customer.

“Customers will notice a difference in what’s being recommended, in the accuracy of what’s being recommended,” Song says. “But it’s subtle. The tech should not be what customers see.

“Where algorithm-based recommendations like Amazon offer you a white shirt if you’ve bought a white shirt before, or if you’ve read a book about Michael Jordan offers another book about Michael Jordan, the AI allows us to predict based on what you haven’t bought in the past and a lot more data. The recommendations are probably going to be a lot more surprising and interesting.”

The retailer oversees the front-end customer experience of the AI service, but works with startup machine learning venture Propulse on the back end. It’s taken six months of testing to get Style Plan’s experience up and running.

“We tested how much data we need to push through the machine learning, how many products we need to show customers… changing the various parameters because it’s a continuous optimization,” says Song. “I don’t know if perfection exists in this space because it’s all about what customers want to see.”

When dealing with customers without a rich purchase history at Frank + Oak, the AI can be given a few starting points by asking the customer to rate various test clothing options it displays. It’s similar to the visual filters being used online by Vancouver’s Shoes.com and The North Face, which launched its AI-enabled digital personal shopper last year.

Such a system “replicates some human processes to assist consumers in digital commerce,” Neil Patil, president of Fluid (the company behind North Face’s AI), told Marketing in 2016. “Where you couldn’t always put a human in front of a shopper, you now can. It gives the [consumer the] ability to have that conversation that they normally would have with a store associate… and they can get what they need 24 hours a day.”