Last week, technology luminary Andrej Karpathy offered a crucial perspective on the AI journey, calling this the “decade of agents,” with truly autonomous, reasoning AI agents still years away.
Many people opined and said his statement sounded pessimistic, offering businesses a temporary reprieve from the urgency of AI.
But that’s not what he was trying to say at all. And for Chief Marketing Officers (CMOs) and Chief Digital Officers (CDOs) in the fashion, home décor, and beauty sectors, the message is one of realism and immense opportunity. No one should view this Decade of Agents as a buffer because it isn’t one. This is not a time to sit back and wait. This is not a time to procrastinate or avoid the inevitable. Rather, we should see this decade as a runway, and get moving now to be ready to take flight in the not-so-distant future.
Agents, in and of themselves, aren’t the breakthrough that will define the next generation of retail. Agents will only perform optimally with the existence of the rock-solid, machine-readable data infrastructure. For brands with intricate product catalogs, rich visuals, and complex size/color variations, the time to build and optimize this data foundation is now. This decade of runway is really a decade of preparation so brands can be relevant and successful in a future when consumers adopt AI search and discovery at scale and also start to trust the safety and authenticity of agentic commerce overall.
Understanding the “Decade of Agents” and Its Retail Context
The core truth in Karpathy’s observation is that today’s Large Language Models (LLMs) are great at generation (writing copy, creating images, etc.) but they are still weak at agency, the ability to act autonomously, deeply reason about price vs. value, fully remember past preferences, understand the nuance of occasion context, and execute a reliable, multi-step purchase decision. Each of these shortcomings is a complex problem in their own right and then made even more complex when trying to combine them with a goal of delivering a unified, smooth consumer experience. These are complex problems and solving the engineering challenge to create more elegant solutions will take time.
Defining Agentic Commerce in Fashion, Home, and Beauty
When agents finally achieve reliable and trusted autonomy, Agentic Commerce will be the new mode of shopping. It may not look like the experiences we are experimenting with today, but there is no doubt we are entering a new era of what digital advertising and commerce will look like in the coming years. And when we eventually get to the point of agentic trust and reliability, we can expect to see customers delegating both the search and the purchase to an AI that acts on their behalf.
We can expect to see personalized agents that will take complex, nuanced instructions and execute our entire shopping journey:
- For Beauty: “Find a clean, cruelty-free foundation that matches my skin and from a reputable US-based retailer that ships within two days.”
- For Home: “Recommend a durable, stain-resistant rug, 8′ x 10′, in a muted botanical pattern that complements the sofa I bought two years ago, and check if it’s compatible with a heated floor.”
- For Apparel: “Show me a formal dress in dark navy, almost black but not black, less than $500.00, available in petite size 4, with 4-star reviews or higher, that I can return in a store near my home.”
The success of the agent, and thus, the sale, is dependent on the agent’s ability to swiftly and accurately understand the consumers’ priorities while simultaneously processing your product data and deciding if it is, in fact, the best choice for a person’s particular needs. If your product information is missing understandable size guides, unclear on material composition, or unclear on occasion and use benefits, then the agent could choose a more relevant item from your competitor.
Product Data Infrastructure Is Your New Competitive Moat
CMOs and CDOs should not view this decade as a waiting period. Instead, they should act now and take advantage of the time they have to methodically rebuild and modernize their product data infrastructure.
The Cost of Waiting: Buffer vs. Runway
- Viewing this decade as a Buffer (Complacency): Assuming you have ten years of time leads to procrastination. When the agent breakthrough accelerates (perhaps unexpectedly in year three, four or five), your brand will be left with unusable, fragmented, incomplete data and a multi-year IT project that should have started yesterday.
- Viewing this decade as a Runway (Strategy): This is the window required to meticulously build and test the systems and processes needed to prepare your products to compete in the new digital advertising and commerce environment. Building agent-ready data is difficult, time-consuming, and must be layered atop existing ERP and PIM systems, so it makes strategic sense to start now, and not waste time.
Three Pillars of Agent-Ready Retail Data
To secure a place in this next era of discovery, retailers and brands must invest in three non-negotiable data pillars:
- Structuring Product Data: This is the move from human-readable text to machine-readable structured data.
- Standardized Attributes: Features like “Waterproof,” “100% Cotton,” “Chocolate Brown,” “Fitted,” or “Made in Italy” must be codified with a single, universally recognized term and meaning, not 15 internal variations.
- Implementing Schema Markup: Using technical SEO practices like Schema.org (Product, Offer, SizeGroup) to explicitly spell out every detail, from available stock to material, in a language AI agents understand.
- Enriching Catalogs with Context: Agents need more than specs; they need context to reason and decide.
- Semantic Value: Enriching product listings to include not just what the product is, but why it matters to the customer (e.g., “This fabric is pilling-resistant and easy-care,” or “This formula includes hydrating Hyaluronic Acid”).
- Policy Clarity: Embedding crucial, policy-related data directly with the product (e.g., return windows, shipping costs, promotional exclusions).
- Making Products Understandable: The Path to Agentic Discovery
- Multi-Modal Clarity: Ensuring that every visual (high-res image, 360-view, style-outfit photo) has clear, machine-readable alt-text and captions that accurately describe its contents and context.
- Q&A Optimization: Utilizing FAQ schema and structured content to preemptively answer the thousands of nuanced questions an agent might ask on a customer’s behalf.
Perfecting What Matters
For Lily, our focus has always been on solving this exact problem: helping brands make their products easier to find, understand, and choose, by people or by AI. We believe that the ability of an external agent to accurately digest your product information is one of the greatest competitive advantages you can build.
The decade ahead is not a time for passive observation. Take advantage of this time, your critical window for strategic investment. The comprehensive, authoritative content you build today must serve two audiences:
- The Human Shopper: Providing the emotional context, visual inspiration, and trustworthy information they need to feel confident in their purchase.
- The AI Agent: Providing the perfectly structured, machine-readable data it needs to successfully execute a complex, criteria-based transaction.
As CMOs and CDOs, your competitive strategy for the next decade must center on this reality: The breakthrough isn’t the agent itself but the data infrastructure that makes the agent work. Progress will come from perfecting what matters—your core product data—not from chasing fleeting AI headlines.