From AI tools to AI infrastructure

In recent years, much of the AI discussion has revolved around tools for text, images, analysis, and productivity.

Commerce is now entering a new phase.

AI is becoming part of the very infrastructure between the customer, the online store, and the products.

Shopify has made agentic commerce a central area in platform development. Shopify Catalog structures and enriches product data so that AI agents can understand products, prices, variants, and availability, and use this information in AI-based purchasing journeys. Shopify makes product data available for AI channels and agents using Catalog.

Shopify has also developed the Universal Commerce Protocol with Google to standardize how AI agents can work with commerce, from product discovery to shopping cart and checkout.

This makes AI more than just a new marketing tool.

It becomes a new technical layer for commerce.

Shopify opens its storefront to AI agents

Development has progressed even further.

From August 5, 2026, Shopify will directly expose WebMCP tools on Liquid storefronts. AI agents can use these for, among other things:

  • searching the product catalog

  • retrieving product information

  • reading and updating the shopping cart

  • proceeding to checkout

  • finding the store's policies and FAQs

This functionality is available on Liquid storefronts without separate installation or configuration. Hydrogen has similar support in developer preview.

This means that the distinction between a traditional online store and an AI-based shopping journey becomes less clear.

The customer no longer necessarily needs to navigate through categories, filters, and product pages in the same way as before. An AI agent can increasingly perform parts of this work on behalf of the customer.

For commerce architecture, this means that structured product data, clear integrations, and control over which systems and data are available become even more important.

At the same time, new requirements apply under the EU AI Act

The EU AI Act is being introduced gradually.

From August 2, 2026, the transparency requirements in Article 50, among others, will apply to certain types of AI systems in the EU. The rules include requirements related to direct interaction between humans and AI systems and the labeling of certain types of AI-generated or manipulated content.

This is particularly relevant when AI moves from internal work tools to the customer experience.

AI-based customer dialogues, assistants, automated content, and new agentic purchasing experiences make it more important to control the actual role of AI in the solution.

The AI Act is also based on a risk-based model. The requirements vary depending on how the AI system is used and the risk its use represents.

For commerce, this means that AI cannot be treated as a single technology category. An internal development assistant, a customer-facing AI assistant, and a system that influences decisions about people can be subject to very different requirements.

Norway follows suit

The AI Act has not yet been implemented into Norwegian law.

The government is working to incorporate the regulation into the EEA Agreement and implement it through a Norwegian AI Act. Following changes to EU regulations, the legislative proposal will undergo a new consultation in autumn 2026, and the government's ambition is to submit the law to the Storting in spring 2027.

At the same time, existing regulations already apply to the use of AI in Norway, including privacy regulations, the Copyright Act, and other technology-neutral legislation.

For Norwegian businesses, this makes AI governance relevant even before the Norwegian AI Act is in place.

AI and GDPR are closely linked

AI also does not change existing requirements for privacy and data processing.

When AI is connected to commerce systems, the solutions can access product data, customer dialogue, order data, support information, or other parts of the company's datasets.

At Appsalon, AI is therefore treated according to the same basic principles as other external services and integrations: control over access, data minimization, and a clear understanding of what information is being processed.

We have internal guidelines for the use of AI. Sensitive credentials, API keys, and unnecessary personal data should not be shared with AI services.

The same principles apply when AI becomes part of a solution we design for clients.

Product data takes on a new role

The development in agentic commerce also makes product data more strategic.

In a traditional online store, products are primarily presented through what the customer sees on the screen.

An AI agent works differently.

It needs structured and consistent data to understand the product's features, price, variants, availability, and relevance.

Shopify Catalog is built precisely for this. Shopify standardizes product data and can, among other things, structure and infer attributes to make products easier for agents to search, compare, and recommend.

Shopify also gives merchants control over which connected AI channels product data can be made available through the agentic part of Shopify.

Thus, the quality of product data, PIM structure, and integrations also become part of a company's AI preparedness.

New interfaces require the same architectural discipline

Shopify's development shows how quickly AI moves from experiment to platform functionality.

AI agents can already work with product catalogs and purchasing journeys, and Shopify is building on open protocols and standards for agentic commerce. The WebMCP support that arrived in August is a concrete example of this development.

For Appsalon, this does not change the fundamental principles of how we build commerce solutions.

AI must be integrated into the existing architecture with control over:

  • data and data quality

  • accesses and authentication

  • integrations and APIs

  • privacy and GDPR

  • security and logging

  • which external services are included in the solution

  • relevant regulatory requirements

AI introduces new opportunities and new interfaces, but the technical solution must still be understandable and controllable.

This becomes increasingly important when both customers and AI agents need to be able to interact with the same commerce platform.