The Customer Isn't Coming To Your Store Anymore—Their Agent Is
Udit Mehrotra is a distinguished product leader who currently serves as Head of Product for North America Languages Experience at Amazon.
gettyMany people working in retail today have already experienced proto-agentic commerce. They just didn’t call it that. When you asked a conversational shopping assistant which of two laptops would suit your needs, or when an AI-powered search interface surfaced a curated shortlist before you’d finished typing your query, that was the beginning. The fully autonomous agent that researches, compares and transacts on your behalf is coming down the road.
My work at Amazon gives me an interesting vantage point on the challenges of building AI systems that serve diverse, multilingual customers. It turns out that they’re a version of the same challenges every retailer will face as agentic commerce scales. Here’s how I think about what’s changing.
Behind the scenes, agents are already deeply embedded in commerce. They power conversational shopping assistants, enrich catalog data in real time and resolve the intent behind ambiguous queries. The companies currently investing in these agents are accumulating data and trust frameworks that will be difficult to replicate quickly.
The customer-facing layer is where we’re lagging, although we’re closer than most people realize. McKinsey analysts estimate that agentic commerce could influence up to $1 trillion in U.S. retail revenue by 2030. The sequence matters: The infrastructure layer is moving faster than the experience layer, which means the window to build foundations is now.
E-commerce has always been designed for human cognition: visual browsing, serendipitous discovery and productive friction. When the primary user is an AI agent, many of those assumptions need to be revisited. The agent isn’t browsing. It’s meeting a goal.
A human browsing your site can be influenced by placement, imagery and merchandising. An agent querying your product data can’t. It cares about structured attributes, accurate availability signals, reliable pricing and compatibility information. If your catalog data is incomplete or inconsistent, the agent may pass on your product and move to a competitor whose data is cleaner.
In my own work building multilingual AI systems, I’ve seen how thin product data creates downstream failures. When a conversational assistant is trying to answer a question about a product in Spanish or French, incomplete attribute data is a product data problem, not a translation problem. The agent can’t surface what it can’t understand. The same dynamic will play out in English as agentic systems scale: Rich, structured, machine-readable product information is the prerequisite for being discovered at all.
A 2026 Braze study found that nearly half of consumers expect to interact with brands through AI agents by the end of this year. The direct brand touchpoint that retailers have spent years building through loyalty programs, email marketing and personalized on-site experiences is at risk of being mediated away.
But loyalty isn’t going away. What’s changing is the mechanism through which it gets expressed. In an agentic world, a customer who prefers a brand will configure their shopping agent to reflect that preference. Brand affinity becomes a parameter in the agent’s decision tree, not a moment at the digital shelf. A customer who tells their agent “always check this retailer first for kitchen appliances” has expressed something more durable than a loyalty card swipe.
The implication is that building genuine brand loyalty, through product quality, pricing transparency and reliable fulfillment, becomes more important, not less. Agents are unimpressed by merchandising. They’re good at evaluating value. Brands that have relied on placement and promotion to drive purchase, without building underlying product quality and customer connection, will feel this acutely.
Two near-term investments pay off regardless of how quickly the agentic transition accelerates. The first is catalog quality. Clean product data, accurate specifications and reliable availability signals are the foundation that agentic systems need to include your products in their recommendations. This work also improves search performance, reduces returns and makes products easier to find through any discovery channel. It’s the clearest right answer across every scenario.
The second is thinking about answer engine optimization alongside SEO. When an AI agent researches a product category, it reads the web with different priorities than a search engine. Structured, precise data and credible third-party reviews matter more than keyword density.
The AI layer will commoditize faster than most people expect. What won’t commoditize is clean, structured, reliable product data and the customer trust it earns over time. The customer isn’t waiting for permission to hand off the shopping. Will you be ready when the handoff happens?
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
