Agentic Commerce Is Here. Your D2C Catalog Isn't Ready.

Agentic commerce is the shift from shoppers searching to AI assistants choosing. When someone asks ChatGPT, Gemini, or Perplexity for "a fragrance-free moisturiser under 800 rupees for sensitive skin," the assistant builds a shortlist and, in a growing number of cases, completes the purchase in the chat. To be on that shortlist, your catalog has to be machine-readable: structured attributes, complete identifiers, live price and stock, and content that ties each product to a real problem. If your data is messy, delayed, or thin, the agent cannot trust it, so it leaves you out.

That is the uncomfortable part. The shopper never sees your storefront, your hero banner, or your launch-day animation. They see a paragraph the assistant wrote, naming three or four products. Either you are in that paragraph or you are invisible, and the thing that decides it is data quality, not design.

What agentic commerce actually means for your store

For a decade the job was to win a click. A shopper typed a query, saw ten blue links, and you fought for the top of the page. Agentic commerce removes the click. The assistant reads across the web, reconciles product data, and hands back an answer with a small set of named recommendations. The candidate set is short, often three to five products, and being "page two" no longer exists. You are either named or you are not.

Two forces make this urgent rather than theoretical. First, the assistants now have transaction rails: in-chat checkout means a recommendation can become an order without the shopper ever opening a browser tab. Second, the traffic is real and growing fast, with AI assistants now one of the fastest-rising referral sources to retail sites. The mistake is treating this as a far-off experiment. The shoppers who use these tools are early, high-intent, and ready to buy, which is exactly the cohort a D2C brand cannot afford to miss.

Why your catalog is probably invisible to AI agents

In the catalogs we audit, the gaps are consistent and unglamorous. None of them are about your brand being unknown. They are about data the agent cannot use. Four show up almost every time:

• Missing identifiers. No GTIN, no consistent SKU, no brand field. An agent matching products across sources needs a stable key. A "Vitamin C Serum 30ml" with no identifier is, to a machine, indistinguishable from forty others, so it gets dropped from the comparison rather than guessed at.
• Stale price and stock. If your feed says in-stock at 1,299 and your site charges 1,499 or shows sold out, the agent learns your data is unreliable and stops surfacing it. Freshness is a trust signal, and a sale price that lags by two days quietly removes you from results during the exact window you most wanted them.
• Thin attributes. The shopper asked for "fragrance-free, under 800, sensitive skin." If your product record does not carry ingredients, claims, size, and price as structured fields, the agent cannot match the query to your product even when it is a perfect fit.
• No problem-to-product content. Your pages describe features. Shoppers ask in problems. If nothing on your site explicitly connects "this serum" to "dullness after sun exposure," the assistant has nothing to reason from, so it reaches for a competitor who wrote that sentence.

Notice that none of these are fixed by a redesign. They are fixed in the layer beneath the storefront, where your product records actually live.

What agent-ready product data looks like

The target is not exotic. Most of it builds on standards your team already half-uses, applied with discipline across every SKU rather than the few you remember to fill in. Agent-ready means each product carries:

• A complete identity: stable SKU, GTIN where one exists, brand, and a clear title that reads the way a person would ask for it.
• Structured attributes: the specifics a query can filter on, such as size, material or ingredients, price, key claims, and compatibility, stored as fields, not buried in a paragraph.
• Live commercial data: price, currency, and availability that match your storefront to the minute, not a feed that refreshes overnight.
• Problem-framed copy: a few sentences per product, and a few comparison or FAQ pages, that name the shopper's situation and explain who the product is for and who it is not.

The real bottleneck is your back office, not your storefront

Here is the trap. Most brands hear "AI shopping" and reach for a new plugin or an "AI visibility" tool bolted onto the front of the site. But a tool can only publish what your records contain. If your price lives in Shopify, your stock lives in a warehouse sheet, and your ingredient list lives in a designer's PDF, no front-end layer can assemble a clean, current, structured product the agent will trust. The data is fragmented, so the output is fragmented.

This is why agent readiness is, underneath, an operations question. A single, governed source of product truth is what feeds a clean, current catalog out to every channel at once, including the AI ones. That is exactly the job of disciplined product information management: one record per product, every attribute filled, one place to update price or a claim so the change reaches the storefront and the agent feed together. Brands that keep stock and price authoritative in their ERP and push them live, for example through a Shopify and Odoo integration, clear the freshness problem as a side effect, because the same real-time data that runs the business also feeds the assistant.

It is the same lesson from the last shift to AI on-site: the model is rarely the hard part, the data feeding it is. We make the same argument about the storefront experience in our work on AI for ecommerce, and it holds here. Clean operations are the unglamorous foundation that makes every AI surface, search, recommendations, and now agents, actually work. For more on how brands are already using these surfaces, our piece on how D2C brands use AI to personalise shopping covers the storefront side of the same coin.

A practical readiness check you can run this week

You do not need a platform migration to start. You need an honest baseline. Pick your ten best-selling SKUs and score each one out of four:

• Identity: does it have a stable SKU, a GTIN where one exists, and a brand field? One point.
• Freshness: do its price and stock match your live storefront right now, with no overnight lag? One point.
• Attributes: are the three or four things a shopper would filter on stored as fields, not prose? One point.
• Problem fit: is there a sentence anywhere that names the shopper's problem this product solves? One point.

Total it up. If your average is below three, you have your roadmap, and it is a data roadmap, not a marketing one. Fix the lowest-scoring dimension across the whole catalog first, because the agent judges your reliability on the weakest link, not the strongest product. A brand we worked with this year found roughly a third of its SKUs missing a usable identifier and a quarter showing stale prices on weekends, when a recurring sale ran. Fixing the source data, not the storefront, was what put those products back into the candidate set.

Frequently asked questions

Is agentic commerce only relevant for big brands?

No, and arguably the opposite. Agentic commerce rewards clean, well-described products, which is something a focused D2C brand can get right faster than a sprawling marketplace seller. It is also a way to reduce dependence on a single marketplace and its fees, because an assistant can send a shopper straight to your checkout when your data earns the recommendation.

Do I need to join a specific protocol or program?

The protocols and merchant programs matter, but they sit on top of the same foundation: complete identifiers, accurate feeds, and structured attributes. Get the product data right and you are ready to plug into whichever channels make sense, rather than rebuilding for each one. Start with the data, then choose the rails.

How is this different from the AI search I already have on my site?

On-site AI search helps a shopper who is already on your store. Agentic commerce decides whether you appear at all, in a conversation happening off your store, often before the shopper has a brand in mind. Both run on the same clean product data, which is why fixing the data pays off twice.

What is the single highest-leverage first step?

Make price and stock authoritative in one system and push them live everywhere, including any AI feed. Freshness is the trust signal agents weigh most heavily, and stale commercial data is the fastest way to get quietly dropped from results.

The short version: agentic commerce does not reward the loudest brand, it rewards the most legible one. Get your product data complete, current, and framed around real problems, fix it at the source rather than the surface, and you turn a channel that is invisible to most stores into one that recommends you by name.