Agentic commerce means AI agents acting on behalf of a consumer or a business in an ecommerce context. Instead of a human navigating a webshop, searching for products, comparing prices, and going through a checkout, an AI agent does all or part of that work autonomously.
Why is it happening now?
Three things have converged. Large language models have become capable enough to understand product catalogs, interpret purchase intent, and follow multi-step transaction flows. Protocols like MCP (Model Context Protocol) and Shopify's Agentic Commerce Protocol make it possible for agents to interact with stores programmatically. And payment providers — Stripe, Klarna, PayPal — have started building support for agent-driven transactions where no human is present at checkout.
The result is that an AI agent can receive the instruction "find a white organic cotton t-shirt, size M, under €40" and actually complete the purchase: searching the catalog, comparing options, selecting a product, entering payment details, and placing the order.
What does it mean for merchants?
For merchants selling via Shopify, Shopware, Norce, or Magento, the challenge is making the store accessible to agents — not just humans. That means several things need to be in place.
Structured product data. Agents read data, not browse pages. Product titles, descriptions, attributes, pricing, availability, and images need to be clean, consistent, and machine-readable. A product page that works well for a human visitor may be incomprehensible to an agent if the data behind it is messy.
Machine-readable APIs. Agents interact with stores through APIs, not through the storefront UI. Catalog search, product detail, cart, and checkout all need to be available as API endpoints with clear documentation and predictable behavior.
Checkout without human interaction. Traditional checkout flows rely on a human clicking buttons, entering addresses, and selecting payment methods. Agent-driven checkout needs to work programmatically — cart creation, address submission, payment method selection, and order placement via API calls.
Clear pricing and availability. Agents compare across stores. If pricing is ambiguous (hidden surcharges, quantity-based pricing that requires manual calculation) or availability is stale, the agent will move on to a store with cleaner data.
Where are we today?
Agentic commerce is real but early. Shopify has launched its agentic commerce protocol, which defines how AI agents can search products, read product data, and initiate purchases via API. OpenAI's Operator can browse the web and complete purchase flows. Google's shopping agents are testing in production. Perplexity has added shopping capabilities that let agents find and buy products directly from search.
On the merchant side, most stores are not yet agent-ready. The biggest blockers are inconsistent product data, checkout flows that require human interaction, and APIs that are incomplete or undocumented. Merchants who have already invested in headless architecture, clean data, and API-first design have a significant head start.
What about SEO and traditional traffic?
This is the strategic question merchants need to address. Today, most ecommerce traffic comes through Google search, paid ads, email, and social. Agents do not use Google the same way humans do. They interact directly with APIs and structured data. If a store is not exposed to agents, it risks becoming invisible to the next generation of buying behavior — even if the store ranks well in traditional search.
This does not mean SEO stops mattering. Human visitors are not going away. But the traffic mix will shift over time, and merchants who only optimize for human navigation may find themselves losing share to competitors who are also optimized for agent access.
How NWT works with agentic commerce
Junipeer already functions as an integration layer between the ecommerce platform and the outside world — ERP, payments, shipping, and product data. The same architecture can expose store data to AI agents via standardized protocols without requiring a new integration for every agent provider.
Frntkey as a headless frontend means the store's presentation layer is already separated from the business logic. This is exactly the architecture agentic commerce requires — agents do not need a frontend; they need clean access to the commerce backend.
We help merchants assess their current agent-readiness, identify the gaps (usually product data and checkout), build the technical layer needed for agent access, and plan the rollout in a way that does not disrupt the existing store operation.
How to prepare
A practical starting point is to audit three things. First, product data: are titles, descriptions, attributes, and pricing clean and consistent enough for a machine to understand? Second, APIs: can an external system search your catalog, read product details, and create a cart via API? Third, checkout: can a purchase be completed without a human clicking through the checkout UI?
If any of these fail, that is where the work starts. The good news is that these are the same investments that improve the store for human visitors too — better data, better APIs, better performance. Agent-readiness and store quality go hand in hand.
See our AI in ecommerce guide for the broader picture of how AI tools fit into the commerce stack, and our platform comparison for how each platform supports API-first and headless architectures.
