Before AI became the default answer to every retail challenge, Purva Gupta was already building Lily AI around a much simpler idea: shoppers and retailers don’t speak the same language.

Now in its tenth year, Lily helps brands rewrite their product data using the language real people use to search. Whether it’s on Google, a retail site, or a generative platform, Lily enriches product catalogs with consumer-first language to ensure items are not just listed, but actually discoverable.

“We’re bridging the gap between how merchants describe products and how consumers describe them,” said Gupta. “If you’re calling it ‘midnight French terry athleisure’ and I’m typing ‘navy hoodie,’ I’ll never find it.”

Tomorrow, Gupta joins the "Beyond the Hype" panel at Possible 2025, where she’ll share what it means to build AI for utility, not headlines.

From Frustration to Foundation

The idea for Lily came from personal experience. After moving to the U.S., Gupta found herself repeatedly frustrated by poor search results while shopping online.

“I thought maybe I didn’t know the right words,” she recalled. But after interviewing thousands of women about how they described their last purchase, she discovered a pattern: what shoppers said and what brands labeled were often worlds apart.

From there, Lily was born. Not in the recent hype cycle, but from a product gap uncovered one conversation at a time.

Gupta, trained as an economist, focused first on the problem, then on building a solution powered by AI. Over the years, Lily evolved to support top brands like Abercrombie, Hoka, and Coach, enriching their entire catalogs with language that maps to billions of real consumer searches.

Focus on What Makes Money

In a crowded AI landscape, Gupta credits Lily’s longevity to one thing: relentless focus on outcomes.

“We’ve stayed competitive by solving the problem that actually matters to brands. And that problem is being found and being sold.”

Lily offers a no-code integration and guarantees a 5 to 10 percent lift in revenue from Google Shopping. For Gupta, the sales lift is the proof, not the model architecture.

“You need the CFO to say yes. They’re looking at ROI, not hype.”

AI as the Baseline

Internally, every team at Lily uses AI in daily operations, from engineering to marketing. But Gupta is clear that AI no longer sets a company apart. It’s simply the new baseline.

“Now the question is how you use your humans,” she said. “AI can get you to 90 percent. Use your team selectively for the final 10.”

That mindset shapes Lily’s product, too. The company offers automated product data generation and optimization, while ensuring oversight when data goes directly to consumers.

Cleaner Data, Less Bias

Since its earliest days, Lily has been built on proprietary training data. That data was carefully designed to represent the language and behavior of the female consumer who Gupta notes controls 80 percent of consumer spending but is often underserved by product labeling.

“Your AI is only as good as your training data,” she said. “And brands are responsible for what goes in front of the customer. That’s why we take full responsibility for what goes into ours.”

Lily combines machine learning, computer vision, and generative tools but it’s the structured, intentional foundation that keeps output aligned and useful.

The Shift to Generative Commerce

Gupta’s team recently conducted consumer research to understand how AI is shaping shopping behavior. The standout stat: four in ten consumers are now starting commerce searches on generative platforms.

This, she said, is not a passing trend. It’s a signal that consumer behavior is evolving faster than brands expect.

“Brands can’t sit on the sidelines. The only way to learn is to test. But the focus has to stay on the fundamentals: what problem are you solving, and what’s the impact of the solution?”

Independently Created. Not affiliated with POSSIBLE.

ClickZ is a ClickZ Media publication in the Events division