Red Van Blog

Agentic Commerce: A 90-Day Readiness Plan

Written by Red Van | Nov 19, 2026, 9:41:34 PM

What agentic commerce readiness really means for your store

Agentic commerce readiness means getting your product data, search, APIs, and operations into shape so AI shopping agents can find, understand, and safely buy your products. In practice, that looks like clean catalog attributes, natural language search, reliable real-time inventory and pricing, and scoped AI support in back-office workflows.

Most commerce leaders already feel the pressure. Your board asks about ChatGPT and shopping agents. Vendors pitch new AI features every week. At the same time, your team is still wrestling with basic issues like inconsistent product attributes, zero-result searches, and brittle integrations. The hype is loud, but the next steps are unclear.

Agentic commerce clears up one thing. AI agents do not browse your site, they read your data. When a shopper asks for “a waterproof rain jacket under $200 that packs small,” the agent compares structured fields such as size, material, use case, price, inventory, and shipping rules. If those fields are missing or inconsistent, you simply will not show up in the shortlist.

The Universal Commerce Protocol (UCP) and related standards make this concrete. The UCP catalog lookup capability provides operations like lookup_catalog and get_product that retrieve products or variants by identifier, returning the full detail needed for purchase decisions. As the spec notes, these operations must resolve product and variant IDs and return structured attributes that agents can understand (UCP catalog lookup spec).

In parallel, the UCP catalog search capability models how an agent will search your catalog using free-text queries and structured filters, with optional context about the buyer. The spec emphasizes that context is a hint, not a guarantee, and that eligibility and policy enforcement happen at checkout (UCP catalog search spec). That means your systems must answer clearly at every step, from search results through cart and order status.

The good news is that you do not need to guess where to invest. Your existing stack already maps neatly to four readiness pillars: data and catalog foundations, AI search and merchandising, API and integration readiness, and AI in operations. The rest of this article turns those pillars into a practical 90-day plan.

Pillar 1–3: Data, search, and APIs agents can trust

Start with the foundation your agents will actually consume: your catalog. Data and catalog foundations mean making sure key information about every product exists, is consistent, and lives in fields a machine can read. That includes titles, attributes, variants, sizing, materials, compatibility, pricing, availability, and policies like shipping and returns.

A simple first pass is to look for five warning signs: details like size or material only in description text, attributes like “Color: Navy” written five different ways, variants missing data the parent product has, pricing or inventory that differ by channel, and policies not stated in any structured field. If you see more than two of those across a sample of products, you have a data problem, not an AI problem.

On Shopify, you already have tools to fix this. Product metafields and the standard taxonomy give you structured homes for attributes such as fit, fabric, or use case. The work is defining an attribute model per product type and mapping it into those fields, then making sure new SKUs cannot go live until required attributes are filled. Even a focused effort on your top 20 percent of revenue-driving products can move the needle.

On Salesforce Commerce Cloud, catalogs are often deeper and more customized, with attributes stored across SFCC, a PIM, and an ERP. Here, the crucial move is agreeing on a single source of truth per attribute and pushing clean data to every channel that reads it. If two systems own “available to promise” for inventory, your agents will inherit those conflicts and misquote stock.

Next, look at how shoppers and agents will find products. A query such as “wide-fit hiking boots for wet trails” should return relevant results, not every SKU that contains the word “boots.” Yet many sites still rely on brittle keyword search. UCP’s catalog search capability is a useful reference point: it supports free-text queries plus filters for category and price, mirroring how real people search and refine (UCP catalog search spec).

On Shopify, that means making better use of native search and discovery before buying another tool. Clean attributes, synonyms, and tuned ranking rules often fix more problems than a new vendor. On SFCC, it may mean deciding whether native search and Einstein are enough or whether a dedicated search platform is warranted, based on your catalog size, merchandising team, and budget.

Finally, agents must trust that your answers are current. API and integration readiness is about making sure systems behind the storefront can return accurate pricing, inventory, shipping, tax, promotions, and order status, in real time. Batch inventory syncs that run nightly, shipping rules hidden in custom storefront code, or order statuses not exposed through APIs are all risks.

Shopify’s own UCP documentation highlights why this matters. Its AI toolkit provides structured commands for agents to search the catalog, build carts, create checkouts, hand off buyers, and track orders (Shopify UCP docs). Every one of those steps calls your APIs. If those calls are slow or inconsistent, the agent will either misquote the order or move on to another merchant.

Pillar 4 and a 90-day roadmap for AI in operations

The final pillar is often the most neglected in strategy decks: AI in operations. The flashiest AI projects sit on the front end, but the most reliable returns are often in repetitive, data-heavy back-office work. The goal is not to replace teams, it is to give them time back on tasks that are easy to check.

Good candidates include drafting product content from a structured template, cleaning catalog data, sorting customer service requests, flagging order exceptions, and surfacing inventory signals. These are high-volume activities with clear success criteria. A human stays in the loop to review and approve, so the risk is controlled while the time savings are real.

Your own AI approach already shows what responsible use can look like. By your internal measures, AI-assisted development has produced roughly 30 to 50 percent faster delivery on net-new code, with 100 percent of that output reviewed and approved by experienced engineers, and zero customer data shared with third-party AI tools. The same principles apply to commerce operations: scoped tools, clear guardrails, and human ownership of the result.

With the four pillars in mind, you can turn agentic commerce from a vague ambition into a concrete, 90-day plan:

  • Days 0–30: Run a catalog audit across a representative slice of SKUs. Score completeness, consistency, and machine readability. Define a standard attribute model per major product type on Shopify or SFCC and document required fields.
  • Days 31–60: Enrich data for your highest-impact SKUs, wire it into metafields or custom attributes, and tighten publishing workflows so incomplete products cannot go live. In parallel, review search logs, set up synonyms, and adjust ranking rules so natural language queries return focused results.
  • Days 61–90: Map the systems an agent-placed order touches, from search through checkout and fulfillment. Identify slow, stale, or fragile integrations, then prioritize fixes that deliver real-time inventory and pricing and expose clear order status APIs. Choose one back-office process, such as product copy or ticket triage, and run a scoped AI pilot with a clear success metric, like hours saved per week.

By the end of that window, you will not have “finished” agentic commerce, but you will have a measurable foundation: cleaner data, smarter search, sturdier APIs, and one validated operational use case. More importantly, you will know where AI earns its keep in your business, and where simpler workflow or integration changes are the better answer.