Accessories seller Stella & Dot is adding a full apparel line to its inventory assortment today after a year of testing the category.

In the mobile app,Stella & Dot’s 30,000 stylists now have the support of algorithms that work to understand their customer pool, including what styles and sizes are most likely to perform, when messages should be sent, and how often they should be reaching out. New  “style boards” serve up recommended outfits that stylists can then share with specific customers via text, or post them en masse on social media or in email blasts.

“Our business has to be simple and enable great customer service,” said Stella & Dot founder and CEO Jessica Herrin. “We see it as a combination of high tech and high touch – we combine the power of the two so that, while there are some things an algorithm will never know about a customer that a stylist would, they’re still getting support where they need it.”

By striking the balance between AI-backed algorithms and personal stylist relationships, Stella & Dot is looking to veer closer toward the Stitch Fix model of new retail and away from the controversial multilevel marketing model that Stella & Dot has been associated with in the past. While its stylists still earn more when they recruit fellow stylists to work beneath them, Herrin said they’re not required to buy a certain amount of inventory at the outset, so they’re not weighed down by masses of product that aren’t going to turn over. Instead, orders from customers are placed by stylists when they’re made, and then stylists earn a certain percentage of commission based on their sales history.

“AI can predict what customers should be shown which product and when,” said Nick Smyth, vp of sales at personalization platform Qubit. “Retailers who master that are the ones who will win, but retailers who can fuse that algorithm base with real-time, human-led personalization have an even better edge.”

While Stella & Dot’s claims around revenue potential for stylists have been called into question for being misleading, the company benefits from its army of stylists that can personalize communication with the end customer on a hyper-local scale. (In response to those claims, Herrin said that many stylists sign up to try it out and receive free product, understanding that you get out of the business what you put in.) Already, Herrin said, sales percentages have increased since the new app was introduced earlier this year, and apparel has boosted that growth. Herrin said that the average value for orders including apparel items is 40 percent more than those that don’t, and she expects that apparel will account for 30 percent of the company’s revenue by the end of this year.

The digital selling tool enabling that growth gives sellers better insight into their customer needs by serving up suggested outfit combinations and telling sellers which customer sets would be most likely to buy those outfits. The algorithm is tracking variables like customers’ purchase history, browsing behavior and demographic, like location and age, on an individual level. The stylist can then add personalized touches, like swapping in a minimal necklace for a statement piece if the customer has a wedding coming up.

As Stella & Dot’s customer-facing algorithms become more mature, the company is looking to them for insight on where to invest in inventory, and expand styles and sizes. Already, in its first year of testing apparel, the company saw that plus-sizes were impossible to keep in stock. In the U.S., it tripled its investment in plus-sizes over straight-sizes as it continued building out its apparel assortment.

Herrin said that the capabilities of the algorithms are still being learned, but that a technology-driven company “roadmap” is coming together.

“We can use this to solve inventory problems, plan the growth of the business and put stylists on the right path to success,” said Herrin. “It’s going to help set our business apart.”