Brands at eTail Palm Springs share lessons on the ‘messy middle’ of building AI tools
When Owen Spencer first started conversations about how artificial intelligence could be used for coding at sporting goods manufacturer Revelyst over a year ago, the first response was apprehension.
This was when ChatGPT was first taking off, and Spencer, Revelyst’s former senior director of DTC applications and AI enablement, said people were concerned AI would replace their jobs if implemented into the business. But now, many of those same engineers are working across more departments and on bigger-picture projects, like helping data analysts on projects or building apps, Spencer said.
“They’ve moved through that moment of unease,” he said. “Initially, it was this feeling of, ‘OK, well if this thing can write code, then my job here is done.’ But they found out really quickly that writing code was just one piece of what they did.”
Spencer was one of the hundreds of brand and retail experts at eTail Palm Springs this week who discussed how AI is transforming their e-commerce operations. Others have included J.Crew, PopSockets and Hibbett. Revelyst owns companies like helmet brand Bell, outdoor gear maker CamelBak and extreme sports gear brand Fox.
Here’s a rundown of lessons brands have shared about their AI implementations so far.
Get teams aligned around AI strategies
Tari Huddleston, global vp of digital and e-commerce at Revelyst, said the company is using AI across almost all areas of the business, except for marketing imagery. But before the tools were rolled out, she said, she spoke with the company’s head of legal to ensure the company could move forward safely.
“AI wasn’t a project. It still isn’t a project. It’s literally a way in which we’re operating going forward,” she said.
Since then, Revelyst has expanded the in-house AI testing and tools it uses across departments. Spencer from Revelyst, who recently left the company to launch a consulting firm called Friends with Robots, said bringing in teams early helped assuage fears and ensure everyone was on the same page. Since then, there have been company-wide announcements, semi-regular training sessions, and a SharePoint drive with AI training and information.
Huddleston said part of the reason the company was able to get its teams on board with AI tools at the same time was that the brands were brought together after Revelyst was formed as a standalone entity in November 2024, following a planned separation from Vista Outdoor.
“We didn’t come together as all brands on the same tech stack,” she said. “We didn’t even come together with the same categories or platforms, or you name it. It’s been a year-long journey just to get everyone on the same platforms and systems. And now we’re able to have that foundation in place, to be able to really figure out how we can leverage AI.”
Get your data clean before giving it to AI
For its part, J.Crew recently rolled out AI summaries of reviews. But before that, it solicited more reviews from past purchasers to increase the amount of data and anecdotes it had from customers about the size and fit of their own products, said Ruchika Jupali, J.Crew’s svp of digital experience and technology.
“We didn’t know we were necessarily doing it for AI,” she said. “And I think that’s just a good premise, in general: The technologies will keep evolving, but you really want to have the best data set and your proprietary data set.”
But then AI became more widely used, and after a series of A/B tests, J.Crew rolled out AI summaries of reviews.
“Had we not invested in that review volume and some of those capabilities, we wouldn’t have actually had good-quality review summaries,” she said. “It’s one thing to have a summary on seven reviews, but it’s another thing to have it on 70 reviews.”
Huddleston from Revelyst also underscored the importance of clean data in getting AI systems off the ground. Something as simple as different naming conventions for files can complicate a process intended to streamline information when it’s fed into an AI system.
“Give your data the time and love it needs,” she said. “Bad data does not get better when AI gets involved. It only gets worse and amplifies it.”
Think function over flash for AI processes
Where AI processes are working best is where they are saving employees time or adding value to the customer experience.
At Revelyst, one key AI win was scraping information from product guides, video tutorials, how-tos and other site content, and putting it in a text field that can be read by AI tools. That optimized the likelihood that the brand would appear in ChatGPT results for customers asking questions relevant to that content. The company worked with the AI platform Perplexity to help scrape the information — a task that would otherwise take a whole team of employees to pull out.
“Our approach has been to look at it from a business perspective and figure out where there’s friction in the organization and how we can start to remove that friction,” Huddleston said. “It’s not just for our consumers on the site. It’s also for our associates sitting in the seats, bringing it all to life for our associates.”
At PopSockets, global vp of e-commerce Jack Farrell said the company has been able to scale the number of Meta ads it runs, resulting in a 50% increase in revenue and profitability last year.
Behind the scenes, the company built a system called BEAM (Breakdown Engine for Ad Metrics) that helps it analyze which, out of the thousands of ads it’s running, is performing best, and also what combinations of copy, image and products are best to use.
“You can’t just tell your designers and your creative team, ‘Hey, I need 500 ads.’ Or, ‘I need 1000 ads.’ ‘I need 2,000,'” he said. “That doesn’t work. You need a system, you need a feedback loop — some systematic approach to this problem.”
Spencer said the industry is in the “messy middle” of implementing AI. While there are proven use cases for AI in chatbots, inventory forecasting and website cleanup, there are still hurdles around getting teams up to speed and figuring out where AI is actually functioning, and not just a flashy use of tech. He said he advises companies to take a breath before acting.
“This is just the beginning, so keep that in mind that, even if you feel behind now, we’re just getting started,” he said. “No matter where you are, just keep on moving forward, be curious, and make sure to take a breath, because it can be overwhelming.”