How to prepare your product catalog with agentic mapping using the Smart Data Modeler

Agentic mapping: Making your product catalog discoverable by AI

With 20% of online shopping estimated to be transacted by AI agents by 2030, brands can’t afford to ignore agentic commerce. The big question every retail company wants answered is: How do you make your products discoverable by AI?

Agentic commerce depends on high-quality product data, and agentic mapping is the process that makes product data readable for AI discovery and checkout.

To recommend a product, AI platforms need clear, structured information, a live data feed and specific product attributes. No doubt, this is a critical step to ready your catalog for agentic sales. But until now, mapping an existing product catalog for agentic channels at scale could be a huge undertaking.

Not anymore. The Smart Data Modeler lowers the barrier for brands and retailers to get their catalogs agentic-ready, available in both commercetools AgenticLift and AI Hub.

Read this guide and discover what agentic mapping is, the best way to map product attributes at scale and how brands can prepare their product catalog for agentic commerce with the Smart Data Modeler.

What is agentic product mapping?

Agentic product mapping, or simply agentic mapping, is the process of standardizing a brand’s product data with the required metadata and attributes for agentic commerce.

Agentic mapping matters because product listings with missing or misnamed attributes won’t simply rank lower — they won’t be recommended at all. Products that don’t meet agentic commerce standards are effectively invisible to AI shopping platforms.

Many brands are already familiar with product mapping for traditional channels like Google Shopping or marketplaces like Amazon. AI platforms have their own schemas, required attributes, naming conventions and rules for products discoverability and checkout. Agentic mapping bridges the gap between your current product structure and the LLM requirements.

What agentic-ready product data looks like

AI shopping platforms need structured, rich, live product data to turn your product into a personalized recommendation. Here’s what brands need to provide:

• Live feeds for pricing, inventory and promotions.
• Consistent attributes and naming conventions, i.e., always “brand” instead of “manufacturer; “Small” instead of “S”.
• Context-focused descriptions like occasion, target audience or use cases.
• Structured data and schema markup.

One important requirement for agentic commerce is mandatory product attributes. OpenAI’s Agentic Commerce Protocol (ACP) requires certain fields for each product, including item_id, title, description, image_url and brand. Many attributes are needed for discovery, while others like your privacy policy and return policy URLs are needed for checkout. Any missing fields will cause a validation issue — your product won’t be recommended or available for checkout.

Other products are highly recommended, like size, color, age_group, dimensions, star_rating, shipping price and return_policy. These help AI chatbots understand your product so they can recommend it to interested shoppers.

How to map product data for agentic commerce

A retailer might have thousands or millions of products with data spread across a PIM, eCommerce platform, ERP, DAM and other systems. Even if the data is rich, that doesn’t mean it’s agent-ready. To map their data manually, a brand would need to:

• Inventory the existing catalog across the PIM, eCommerce platform and other systems.
• Review the target AI platform’s specifications and create a crosswalk between those requirements and existing product fields.
• Determine the source of truth for each attribute.
• Resolve inconsistencies in naming, formats, values and variant structures.
• Identify missing mandatory fields and determine how to populate them.
• Build transformation logic to convert the existing catalog into the required structure.
• Create the new product schema in the commerce platform.
• Transform and import the catalog.
• Repeat the exercise as AI platforms change their requirements or the retailer adds another channel.

This exercise is particularly difficult at scale. A retailer with 100,000 products isn’t just mapping 100,000 records — it may be mapping hundreds of thousands or millions of variants and attributes. This transformation isn’t a one-time fix. Retailers need a data model that can pivot across traditional and AI channels without rebuilding their catalog for each.

Instead of pointing to a manual solution at an AI challenge, we decided to create an AI-powered solution to attack this problem.

How to model product data with Smart Data Modeler

commercetools’ Smart Data Modeler is an AI-powered assistant that eliminates the guesswork and manual legwork of product data modeling. The tool analyzes a brand’s existing catalog, instantly identifies structure, attributes and variant patterns and updates them in the system with your input. Here’s how it works.

Instead of a full manual audit, you start with a representative export from your existing system (either csv, xlsx, xls or json format). The goal of this step is to provide Smart Data Modeler with sufficient representative data to understand your existing structure.

Before uploading, check important fundamentals such as brand, condition, gtin, image_link and description, as well as confirming that product variants have the correct grouping logic. The tool with prompt you to select your target agentic channels from OpenAI, Stripe Agentic Commerce Suite or both.

Smart Data Modeler analyzes the sample and generates a proposed data structure for your selected agentic channels, in three categories:

• Attributes mapped with platform-ready fields: Smart Data Moderler shows characteristics from your file mapped to the corresponding platform attribute, with attribute type, level (product or variant) and any required actions (e.g., approve, rename).
• Attributes not needed in commercetools: Fields from your catalog that AI shopping platforms can’t display in product feeds. You can add these later if you expand to non-agentic channels.
• Already defined fields in commercetools: Smart Data Modeler recognizes native commercetools product fields such as sku, name, description, images, categories, slug and key, avoiding duplicate work.

The human in the loop still has an important role: Reviewing the mapped fields and resolving or approving exceptions.

Smart Data Modeler flags situations where:

• Multiple catalog fields could map to one platform attribute.
• A mandatory platform attribute is missing.
• A decision is needed about which existing field should take priority.
• A default value may need to be applied.

For example, if a retailer has Product Color, Primary Color and Display Color, Smart Data Modeler can identify that these potentially map to the same destination attribute. The retailer can then decide whether to merge them, prioritize one source or use a single field.

If a missing value is identified with no match in your system, you can add a default value that applies to all products.

Once you approve the mapping, Smart Data Modeler can create the Product Type and Attributes and transform the product catalog into the new data model right in commercetools — no exporting and importing required.

If you want to handle the data transformation outside of commercetools, you can download the resulting JSON and import it through the commercetools HTTP API, or create the attribute schema as a template.

Once the product transformation is complete, you can continue configuring your agentic channels to sync the products to the relevant AI platforms.

Why agentic product mapping is essential for retailers

As consumers send billions of AI prompts a day, retailers need agentic mapping to keep their products relevant and visible in online search. But beyond customer reach, agentic mapping helps with a growing operational challenge.

While agentic mapping may seem like a technical checklist item, it’s part of something much larger: A product data strategy. Retailers are grappling with hundreds of thousands of variants, multiple external data sources and now many channels each requiring their own sets of standards. Brands need a roadmap to translate these complex, inconsistent product models into structures that AI channels can understand.

Product data modeling with a flexible, headless platform like commercetools lets brands create a single source of truth and structure their product catalog for different channels.

How to get started with commercetools’ Smart Data Modeler

Agentic mapping with Smart Data Modeler is available to all AI Hub and AgenticLift customers on commercetools:

• AI Hub enables brands to sell on AI channels and manage the logic behind discovery and transactions, all in one place.
• AgenticLift allows brands to transform their product data for agentic commerce without replatforming.

At the moment, customers need to submit a support request with their Project Key to enable Smart Data Modeler.

Once enabled, brands can access it by visiting: Merchant Center → Settings → Product types and attributes → Add product type → choose Agentic channels → Continue to Smart Data Modeler.

Ready to learn more? Request a demo of AI Hub or AgenticLift.