Agentic commerce readiness: prepare your store for AI shopping agents

This guide is for ecommerce managers, CTOs and founders of mid-market retail, D2C, marketplace and print businesses who keep hearing that "AI agents will shop for customers" and want to know what to do about it now. It explains the three open protocols published so far, gives a readiness checklist, looks at the hardest case, personalised and configurable products, and lists the risks to control. It is part of our technology trends 2026 series.

In this guide

• What agentic commerce is
• The open protocols so far
• Why it matters to mid-market merchants
• The readiness checklist
• The hard case: personalised and configurable products
• Risks and how to control them
• A ninety-day start
• Where Netbase stands
• Limitations of this guide
• Frequently asked questions
• How this guide was made
• Next step

What agentic commerce is

Today a shopper visits a store, reads pages and clicks through checkout. In agentic commerce, the shopper tells an AI assistant what they want, "a pack of 500 matte business cards, delivered by Friday, under a set budget", and the agent does the finding, comparing and ordering, asking the shopper to approve at the key points.

That changes who reads your store. The agent does not see your hero banner or your persuasive copy first. It reads structured data: product attributes, variants, prices, stock, delivery options, return rules and the checkout interface. If those are incomplete or inconsistent, the agent either skips your product or orders the wrong thing.

For merchants the change arrives through the channels shoppers already use: AI assistants, search engines with AI modes and the commerce platforms that host stores. Nothing has to be rebuilt overnight, but the data and interfaces that agents depend on take months to put right.

The open protocols so far

The protocols overlap and will change. The lesson for a merchant is not to bet on one of them, but to make the underlying data and commerce functions clean enough that connecting to any of them is an integration project, not a rebuild.

Why it matters to mid-market merchants

Discovery moves upstream. When an agent shortlists products, a store with poor attributes is invisible, however good its design.

Checkout becomes an API. An agent-led order still needs your pricing rules, tax, shipping and payment logic. If those live only inside page templates, an agent cannot use them.

Trust decides conversion. Shoppers will approve agent purchases from merchants whose prices, stock and delivery promises prove accurate. Errors that a human would notice and forgive become failed orders.

APIs meet a new kind of caller. Postman's 2025 State of the API report found that 51% of developers name unauthorised or excessive API calls from AI agents as their top security concern. Agent traffic needs rate limits, scopes and monitoring from the start.

The hard case: personalised and configurable products

Standard products with a barcode are the easy case. Many mid-market merchants sell something harder: print products with paper, finish and quantity options, custom apparel with print zones and sizes, configurable furniture or B2B items with negotiated prices. An agent cannot buy what the catalogue cannot describe.

Three things make these products agent-ready:

• A configuration model in data. Options, dependencies and price rules are defined as a model the storefront and an API can both use, not hard-coded in a designer tool.
• Artwork and proof steps an agent can hand over. Uploading a file, checking it and approving a proof usually need the shopper. The agent should prepare the order and pass the creative step back.
• Templates for repeat orders. Reorders of known designs, such as the same business cards with a new name, are where agents add value first.

This is familiar ground for Netbase. We have delivered 50+ custom web-to-print platforms, product personalization with online designers and configurators, and we run Printcart, a web-to-print and print-on-demand platform, and CMSmart, whose product customizers 7,000+ businesses have used to launch stores. For 4over4, our conversion work, which included a recommendation engine and automated Adobe Illustrator to SVG file conversion, lifted revenue 82% within six months. See how a web-to-print platform and product personalization fit together.

Marketplaces add a further layer: many sellers, shared catalogues and split payments. An agent comparing sellers needs consistent attributes across all of them; our marketplace platform architecture guide and the ecommerce marketplace solution cover that model.

Risks and how to control them

• Wrong orders. Validate every agent order against the configuration model and price rules, and confirm the summary with the shopper before payment.
• Fraud and abuse. Treat agents as a new client type: authenticate them, scope what they may do and rate-limit calls.
• Manipulated inputs. Product text, reviews and uploaded files can carry instructions aimed at AI agents. Keep instructions out of data fields and give agents only the actions they need, as OWASP's guidance on excessive agency recommends.
• Pricing and promotion errors. Keep one pricing service for the storefront and the API, so an agent never sees a different price.
• Returns and disputes. Record the authorisation and the agent's actions, so a disputed order can be traced.
• Governance gaps. Decide who owns agent-channel decisions and how they are reviewed; our AI governance starter for mid-market companies sets out a baseline.

A ninety-day start

• Days 1 to 30. Audit product data for the top 20% of products by revenue: attributes, variants, prices, stock and policies. List the gaps.
• Days 31 to 60. Fix the data model and publish documented read APIs for catalogue, price and stock. Add rate limits and logging.
• Days 61 to 90. Prototype one agent-assisted journey, such as reordering a known product, with human approval before payment, and measure its accuracy.

Where Netbase stands

Netbase has delivered commerce on WooCommerce, Magento 2, Laravel and headless commerce, including a headless multi-store reward shop with a points-plus-cash checkout for a Dubai-based loyalty and rewards technology company, and it runs its own commerce products. Agentic commerce is a growth capability for us: we are investing in it, we work with the major commercial and open-weight AI models chosen per project, and we have not yet published a delivered agentic checkout. Our ecommerce development service is where that work runs.

Limitations of this guide

• The protocols are young and change often; check each publisher's current documentation before you build.
• Protocol descriptions summarise the publishers' announcements; they are not endorsements of any protocol or vendor.
• Legal questions about agent-authorised payments, consumer rights and liability vary by country. Take legal advice.
• The checklist is a starting point for mid-market stores, not a certification or a guarantee of agent traffic.

How this guide was made

The Netbase Editorial Team wrote this guide from the protocol announcements of Stripe, Google and Shopify, public API and security research, and Netbase's verified commerce facts. David (CEO) reviewed every Netbase statement. Drafting used AI assistance (Claude).

Next step

Send us your top products and your current platform, and we will book a solution review to score your agent readiness. You can also read our custom ecommerce platform guide, see where retail and ecommerce is heading, or browse more Netbase insights.

Related services and solutions