I've spent years designing search interfaces for e-commerce. Filters. Facets. Category trees. I watched users click through 12 tabs, compare specs across three sites, and still leave unsure if the lens they wanted was compatible with their camera body.
I knew the problem. We all knew. But the tools we had—static pages, predefined filters, rigid category hierarchies—couldn't solve it. We built interfaces for databases because that's what technology allowed. Users were trying to solve problems: "I want to start a YouTube channel. What gear do I need?" Yet our interfaces forced them to become camera experts first.
AI agents change what's possible. This is the problem AI agents promise to solve. But here's what most product discussions miss: “The interface isn't the product. The reasoning is”
I built ShopSense AI—a proof-of-concept agent for consumer electronics—to understand this deeply. Not to ship a feature, but to understand how agents think about helping people find the right gear. Github: link
ShopSense AI is a ReAct agent for consumer electronics discovery, built on the SAP Spartacus electronics sample catalog (via API Mock):
The brand positioning: "Find your perfect camera, intelligently matched."
This isn't B2B procurement. This is a consumer navigating a complex purchase decision with constraints they may not fully understand yet.
Traditional product thinking optimizes the path from A to B. Reduce clicks. Streamline flows. AI agents invert this:
| Traditional Approach | Agent-Based Approach |
|---|---|
| Design the interface first | Understand intent first |
| User learns the catalog structure | Agent learns what the user is trying to accomplish |
| Static category pages | Generative, context-aware layouts |
| Comparison tables for everything | Comparison only when user asks |
ShopSense has two parts:
Most demos focus on the second part. They show chatbots with product carousels. What they miss is the first part—the reasoning—and that's where the product value actually lives.