We’ve all had experiences like these: You search for a product online, maybe a new pair of running shoes, but one click turns into a spiral.
Before long, you’re wading through hundreds of results—styles you’d never wear, kids’ sneakers (even though you’re an adult), and options that don’t match your budget. When you’re buried in junk, having more choices doesn’t actually feel helpful.
Rather than shout “Here is EVERYTHING,” AI has the capacity to create more guided experiences, closer to working with a helpful in-store associate. In many cases, though, it’s not quite there yet.
AI IS EVERYWHERE
And yet, expectations are rising.
That’s because for a growing number of people, AI is becoming a default interface. People use generative AI tools every day—asking questions, planning trips, troubleshooting problems, and making decisions.
According to data from Constructor and Shopify, nearly two-thirds of people have used tools like ChatGPT in their daily lives, up from 29% in 2023. Among Gen Z, that number is even higher, with 78% having used GenAI.
It’s only natural that those behaviors and comfort levels carry over into shopping—to the point that today, people aren’t asking “Should AI be a part of shopping?” but rather “Why isn’t it better yet?”
WE’RE EARLY ON
The reality is, we’re still in the first inning of AI in shopping. In particular, when it comes to using AI to help find products, it’s a decision problem instead of a language problem.
In other words, today’s AI systems can understand and respond to complex, natural-language queries like “I’m planning a tailgate, what do I need?” or “Help me find new running shoes.” Several years ago, those questions wouldn’t even make sense to type in a search bar. Today, shoppers can get recommendations that make sense.
The larger and more pressing issue is whether the recommendations make sense for them. That’s where the decision problem lies, because understanding what to show each shopper is tough. It requires detective work, since people’s decisions are often rooted in their prior actions, preferences, behaviors, and so on.
While today’s large language models excel at generating answers—often very confidently—they can struggle with connecting those answers to real-world outcomes and context, like: Which pair of running shoes will make this shopper most likely to buy?
WHY THE GAP EXISTS
To truly help shoppers, AI needs to understand what makes them tick. But general-purpose agents like ChatGPT and Claude don’t have access to important clues: what you bought, almost chose, returned, etc. This information is fragmented, spread across retailers’ systems and it is often proprietary.
But it’s critical to getting the full picture. And without that picture, AI struggles to narrow down what fits your needs specifically.
Like with the running shoes: A serious runner might care more about stability, toe box width, and whether a shoe is better for trails or roads. They might prefer a certain brand or have really liked the last version of a particular shoe. A more casual runner may just want something comfortable for occasional jogs.
So, an “Ask me anything” approach—“What are good running shoes?”—often fails to connect the dots. And if shoppers have to explain every preference and use case themselves, then AI isn’t really simplifying their experience.
Instead, AI needs the right data and context at the right moment to help shoppers make their decisions.
EARLY TRACTION
A context-based approach is showing promise. For example, some retailers have launched their own agents combining their product and inventory data with shopper information, like real-time behavior on site, past purchases, and loyalty status.
So, when someone asks for guidance, the AI can move beyond generic recommendations, showing items that person will likely want.
Not everyone wants to interact this way, and engagement is still early. But even with a relatively small number of people using these types of tools, the impact appears meaningful:
- Amazon shared that shoppers who consult its AI shopping assistant are more than60% more likely to complete a purchase during their session. Usage is rising too, with engagement up almost 400% year-over-year.
- Walmart has seen similar trends: Customers who use its Sparky AI have an average order value that’s 35% higher than other shoppers.
- On sites with AI agents during last year’s shopping period that ran Black Friday through Cyber Monday, more than 10% of revenue came from shoppers who used them, according to our data.
Not everyone has mastered context yet, though: I spent time on a national department store site the other day, adding four pairs of shoes to my cart. The next day, I returned, asking the site’s AI agent to recommend styles similar to what I’d been browsing. The response: “To help me narrow this down, were you looking for men’s or women’s shoes?”
I’ll say it again: It’s early, and there’s a lot of experimentation going on. Retailers are trying to figure out where, with the right context, conversational agents add the most value. So far, areas of high intent, like retail search bars and chat, seem to work well.
Key moments of decision, like on product pages, are another fit. At that point, shoppers often need answers to a few lingering questions like, “Do these run true-to-size?” or “Are these shoes good for wide feet?”
WHAT’S NEXT
As context improves, we can expect AI to become a more useful and prevalent shopping companion.
We’re also likely to see more empowered interfaces, ones that don’t just infer our preferences, but ask clarifying questions as they learn and adapt. There will be a shift, too, from answering to acting, with agents guiding choices more directly and helping with next steps.
With all these advancements, the future of AI in shopping will be defined by how well they understand context and help people act. Then, AI will help you walk away confident in your purchase.
Kevin Laymoun is chief customer officer and chief revenue officer at Constructor.