Macy’s simply launched an AI-powered purchasing assistant. Customers who use it spend almost 400% extra

The chatbot launched across all the company’s digital platforms on Monday, but it was tested with about half of Macy’s website visitors over several weeks, the company told Bloomberg. Shoppers who use the chatbot spend about 4.75 times more than those who don’t, Bloomberg reported.

The bot’s short-term success comes as Macy’s tries to make its comeback after a decade of declining sales. 

Earlier this month, the company reported net sales decreased by 2.4% last year, but returned to comparable sales growth, up 1.5%. Macy’s expects to make $21.4 billion to $21.65 billion in net sales this year, a little less than last year’s $21.76 billion, and sees comp sales flat at the midpoint of guidance. 

Chief Customer and Digital Officer Max Magni explained that customers may be primed to spend more because they’re looking for a specific item, such as an outfit for an upcoming event, rather than when they’re just browsing, Bloomberg reported. He suspects that the bot is also attracting a younger customer base.  

The most popular features are the “complete the look” option, where the bot suggests accessories to go with an outfit, and a virtual try-on feature that allows shoppers to see what an item looks like on them. Customers can also use the virtual try-on feature in store, if they don’t have time to see if an item fits, Chief Stores Office Barbie Cameron told Bloomberg. 

More AI shopping assistants are coming as companies and startups bet on making online shopping more seamless. For example, Bill Gates’s daughter Phoebe Gates founded Phia, a browser extension that compares prices across the internet. 

And after more than four years in beta, Marc Lore and Melissa Bridgeford, publicly launched shopping agent Wizard in February. 

“Every retailer is trying to figure it out one step at a time,” Magni told Bloomberg. “This is anybody’s game. Nobody has cracked the code.”

Getting the Macy’s bot ready for customers has taken some tweaking, and thousands of employees weighed in, according to Magni. Originally, it didn’t take into account that shoppers in different climates may not want to see the same selections. 

There were also some tone issues, Magni added. When he asked for T-shirt suggestions for his son, the bot coldly offered a list and wrote: “Here’s a T-shirt for a 10-year-old.”

Now, the bot is more friendly. When asked again, the bot replied “‘Ten-year-olds can have so much fun with color – do you want a brighter or more muted color selection?’” Magni said. “The machine continues to learn.”