AI-powered shopping agents are reshaping how people discover and evaluate products. With natural language interfaces and personalization capabilities, they promise a smoother, smarter shopping experience. But under the surface, many agents still rely on legacy ecommerce business models—models built for a very different digital environment.
Whether they’re guiding users to external brand websites through affiliate links or returning ranked lists of product URLs like traditional search engines, most current agents don’t yet fully own the shopping journey. The experience has evolved, but the economics and mechanics remain rooted in the past.
1. Affiliate Redirection: A Familiar Model in a New Form
Many agents today provide a chat-based experience that feels personalized and intuitive. They suggest products, answer questions, and help narrow down choices. But when it comes to completing the purchase, they typically redirect users to a brand’s or retailer’s website—just like a blog post or price-comparison site would.
This is the classic affiliate model: the agent earns a commission when users click through and buy. It’s low-risk and easy to implement, but it limits control over the end-to-end experience, from checkout to support.
2. AI Search: The Old Search Engine Model, Repackaged
A second approach mimics Google-style search, just with conversational AI. Instead of personalized recommendations or integrated shopping, the agent returns a list of links (sometimes with thumbnails and summaries) for the user to browse and decide.
This model doesn't guide decisions or facilitate transactions. It simply reorders and rephrases search results—still pushing users out to retailer websites. It’s helpful for certain research-heavy queries but offers limited value for shoppers looking for curated guidance or a streamlined path to purchase.
3. Experience vs. Outcome
Both of these models—affiliate redirection and link-based AI search—prioritize presentation over control. The agent serves as a gateway, not a destination. As a result:
-
The agent doesn’t control inventory or availability
-
The checkout experience happens elsewhere
-
Customer data and feedback are lost post-click
-
The user journey is fragmented across multiple systems
This limits how much value the agent can provide and how much it can learn from user behavior.
4. What True Transactional Agents Could Unlock
Some emerging platforms are beginning to build “closed-loop” systems, where the agent stays with the user throughout the journey: from discovery to selection to checkout and post-purchase support. These agents can:
-
Access real-time inventory and store-level data
-
Offer in-app or in-conversation checkout
-
Bundle products from multiple sources
-
Handle returns and follow-up service
-
Capture rich behavioral and preference data
Instead of acting as a referral engine, they function more like a digital store—with guidance, execution, and accountability all in one place.
5. Why Legacy Models Persist (for Now)
It’s not surprising that most agents still rely on older models. Integrating real-time inventory, payments, fulfillment, and customer support requires deep partnerships, infrastructure, and operational complexity. Starting with an affiliate or search model is faster and requires less overhead.
But those models are also easier to copy—and offer thinner margins and weaker moats. As consumers grow more comfortable with AI agents, the competitive edge will come from owning the experience, not redirecting it.
A Moment of Transition
We’re in a transitional phase.
Conversational commerce is advancing quickly, but many agents are still anchored in legacy monetization frameworks. For the next generation of shopping agents to reach their full potential, they’ll need to go beyond simply listing or linking.
Future shopping agents will need to guide, transact, and support. That’s where both consumer value and sustainable business models lie. This is a transformational shift and our work in getting to this state should not be hindered too much by incremental progress.