Skip to content
Motillo

AI-first e-commerce platform – and how agentic commerce is changing e-commerce

Jakob Twedmark

Jakob Twedmark

VD & Rådgivare

AI-first e-commerce platform – and how agentic commerce is changing e-commerce

Almost every e-commerce platform calls itself AI-first these days. Often, that means a text generator for product descriptions built into the admin panel. That can be useful, but it is not where the big shift is happening.

AI is changing three things at once: how e-commerce is built, how it is run and how customers find and buy. The last of these is called agentic commerce: AI agents that search, compare and increasingly also buy on the consumer's behalf. In this article, we explain what an AI-first e-commerce platform is, what is hype and what you should demand when choosing a platform.

E-commerce platform

What is an AI-first e-commerce platform?

An AI-first e-commerce platform has AI built into its workflows, not just bolted on as an add-on. In practice, this happens on three levels:

  1. AI in day-to-day work. Product copy, translations, image captions, metadata and landing pages are produced with AI and reviewed by a human before publishing.
  2. AI in development. The platform is built so that AI coding assistants and agents can understand, extend and maintain it. This shortens the time from idea to feature.
  3. AI-ready for new buying channels. Product data, APIs and structure make the store readable and purchasable for AI assistants such as ChatGPT, Gemini and Perplexity.

Most platforms have level 1. The difference shows at levels 2 and 3.

A man sitting in front of a computer
Agentic commerce

What is agentic commerce?

Agentic commerce means that an AI agent carries out parts of the buying journey on the customer's behalf. The agent can search for products, compare alternatives, fill a shopping cart and, in some cases, complete the purchase.

This is no longer just an idea. Since 2025, the major tech companies have launched open protocols for exactly this:

  • Agentic Commerce Protocol (ACP) from OpenAI and Stripe (September 2025). Lets AI agents complete purchases without seeing the customer's card details.
  • Agent Payments Protocol (AP2) from Google (September 2025). A way to prove that the customer has actually approved a purchase made by an agent.
  • Universal Commerce Protocol (UCP) from Google together with Shopify, Etsy, Walmart and Target, among others (January 2026). Covers the entire chain: product discovery, cart, checkout and order management.

Progress is not linear. OpenAI, for example, shut down its first checkout solution in ChatGPT (Instant Checkout) in March 2026 after weak sales. The focus has instead shifted to product search and discovery in ChatGPT, with ACP as the underlying infrastructure. Interest is still growing fast: in Sweden, searches for “agentic commerce” have gone from practically zero to over 200 a month in a year.

What does this mean for European merchants today? Agentic purchases are still mainly an American phenomenon. But discovery is already happening here: customers ask AI assistants which product to choose, and the answers are based on the data and content the AI can read. Merchants who have structured product data and open APIs in place when checkout features reach Europe will have a head start.

Changes in e-commerce

Five ways AI is changing e-commerce right now

  1. More efficient content work
    Product copy, category texts and translations for new markets take days instead of weeks. The key is the workflow: AI writes drafts, a human approves them. Otherwise, you risk generic content that neither ranks nor converts.
  2. Visibility in AI search (GEO)
    Generative engine optimization (GEO) is about being cited and recommended in AI answers, not just ranking in Google. It requires clear structure, facts, FAQs and structured data. Read more about our SEO & AI optimization.
  3. Faster development
    With AI coding assistants, building new features, integrations and campaign pages is significantly faster, provided the platform is open and well documented. Closed platforms with proprietary template languages and limited APIs benefit less from AI.
  4. Personalization and search
    AI-based search understands what the customer means (“warm jacket for bike commuting”) instead of just matching words. Recommendations can be adapted to behavior and context.
  5. Purchases via AI agents
    As described above: the store needs to be readable and, over time, purchasable for agents, through APIs and protocols such as ACP and UCP.
Two men talking in front of a computer
E-commerce platform

Why architecture matters: open or closed platform

To get the full value of AI, the platform needs to be:

  • API-first/headless, so that data and functions can be accessed by your own frontend, AI tools and external agents alike.
  • Open and extensible, so that you can plug in new AI services and protocols as they emerge, without waiting for the vendor's roadmap.
  • Free from lock-in, since the field is changing fast. The best tool today may not be the best a year from now.

That is why we find that open-source platforms like Medusa are particularly well suited for AI-first e-commerce. The code is open, modular and written in TypeScript, a language AI coding assistants handle very well. Medusa also has its own tools for AI-first development, such as an MCP server and a CLI.

Checklist

Checklist: what to demand from an AI-first e-commerce platform

Human review. Is AI-generated content saved as drafts for an editor to approve?

Multilingual support. Can AI translate for new markets directly in the workflow?

SEO and GEO. Is there support for structured data, metadata and analysis of visibility in AI search?

Open APIs. Can AI agents and external tools access product data, inventory and prices?

Developer-friendliness. Can your team (and AI assistants) extend the platform freely?

No lock-in. Do you own the code and the data? Can you switch AI services or partners?

Pricing model. Does the platform take a percentage of your sales or charge extra for AI features?

A woman using a computer and looking at a screen
Motillo Pegasus

Example: Motillo Pegasus

Motillo Pegasus is our AI-first, license-free e-commerce platform built on Medusa. It is based on the same principles as the checklist:

  • Motillo AI writes product copy, translates, performs SEO/GEO analyses and drafts CMS pages. Everything is saved as drafts for review.
  • Search-optimized landing pages are just a prompt away, with the CMS of your choice (e.g. Storyblok, Sanity or Payload).
  • Open source with no license or transaction fees, so you own the platform.
  • Ready for B2B and B2C, with integrations for ERP systems, payments and email marketing.
  • Fixed-price migration in 6–12 weeks.

Book a demo of Pegasus →

Summary

An AI-first e-commerce platform is more than a text generator. It makes content production faster and development faster, and it prepares the store for a future where an ever-larger share of the buying journey happens through AI assistants. Agentic commerce is still early, especially in Europe, but the protocols already exist and discovery is already happening. The smartest choice today is an open platform that can keep up with developments without locking you in.

Frequently asked questions about AI-first e-commerce platforms

  • An e-commerce platform where AI is built into the workflows for content, development and search, and where data and APIs make the store accessible to AI assistants and agents.

  • Agentic commerce means that AI agents search, compare and complete purchases on the customer's behalf, using protocols such as ACP, AP2 and UCP.

  • Customers already use ChatGPT and other AI assistants to find and compare products. The ability to complete purchases directly within the AI service is so far mainly available in the US.

  • Generative engine optimization: optimizing content and data so that your brand and products are cited and recommended in AI-generated answers.

  • Many clients get the benefits of AI because their partner uses AI tools and modern ways of working to work faster and more efficiently. In that case, the relationship between client and agency does not change.

    At the same time, some clients want to be involved in development but also want a reliable partner who is available with experience and technical advice, and who takes responsibility for quality and operational reliability. For them, a hybrid model together with a partner can be the right approach.

    Clients with strong in-house technical expertise who are ready to take responsibility for the solution themselves can certainly work independently without external help.

  • Open code and open APIs make it possible to plug in new AI tools and protocols directly and let AI coding assistants work efficiently in the codebase. You don't have to wait for a vendor's roadmap.

More articles

Got any questions?

Get in touch and we'll talk about your growth journey

Jakob Twedmark

Jakob Twedmark

CEO & Advisor

+46 73-074 06 31jakob.twedmark@motillo.se