McKinsey estimates that this shift could orchestrate $3 trillion to $5 trillion in global retail revenue by 2030, with up to $1 trillion of that in the US B2C market alone. Whatever the exact UK figure turns out to be, the direction of travel is clear enough that waiting for certainty isn't really an option. This piece explains what's actually happening, where the real risks sit for online stores, and what you can practically do about it now, not in three years.

What Is Agentic Commerce and How Is It Different From Automated Recommendations?

Agentic commerce describes online shopping where an AI agent acts on a customer's behalf rather than alongside them. It doesn't just suggest a product and step back. Given a goal and a budget, it retrieves information, weighs options against stated preferences, and can complete the transaction itself, under authority the shopper delegated earlier.

That's a genuinely different category from a chatbot or a recommendation engine, both of which still leave the final click with a person. The distinction that matters for merchants is execution. A recommendation system narrows choices. An agent can close the sale. Once software is authorised to complete a purchase without a human confirming each step, your store is no longer only optimising for people; it's also being evaluated, compared, and selected by machines acting on someone else's instructions.

This changes the job of a website in a fairly fundamental way. A product page written to persuade a browsing human is not automatically legible to an agent trying to extract price, availability, delivery terms, and return policy in a structured, unambiguous format. If that data is buried in marketing copy or missing entirely, the agent may simply move on to a competitor whose feed is easier to parse.

How Does an Agentic Commerce Protocol Actually Work?

An agentic commerce protocol is a technical standard that lets an AI agent prove who it is, prove it has genuine authority from the shopper, and complete a transaction in a way payment systems can verify and later audit if something goes wrong. Without a shared protocol, there's no reliable way for a merchant's checkout to distinguish a legitimate agent transaction from a fraudulent one, or for a card issuer to know which party actually approved the spend.

There isn't one dominant protocol yet, and that's part of the difficulty. The major card schemes have each built their own agent authentication frameworks around tokenised credentials and verified identity, while a separate layer of web-native protocols is emerging directly from AI platforms and payment providers, covering everything from delegated purchase intent to machine-to-machine micropayments. These approaches overlap in places and diverge in others, and none has been formally adopted as a UK standard.

For a merchant, the practical implication isn't "pick a protocol and integrate now." Most UK stores can't, since the infrastructure to test against isn't widely live here yet. The more useful move is building the data and policy foundations that will work no matter which protocol eventually wins.

Why Is Liability Still a Grey Area for UK Retailers?

This is the question that should concern any finance or risk team thinking about AI-led buying seriously. If an AI agent completes a purchase that turns out to be unauthorised, mistaken, or fraudulent, who bears the cost? Existing UK consumer protection law applies regardless of whether a human or an AI made the buying decision, but that principle doesn't settle which party in the chain- the merchant, the payment provider, the AI vendor, or the consumer- ultimately absorbs a disputed transaction.

Card schemes are starting to build purchase protection mechanisms for agent-initiated transactions, and UK financial regulators are actively reviewing how existing frameworks apply to AI-driven payments. None of that has produced a settled legal position yet. Until it does, the strongest position belongs to businesses that have already written clear internal policies covering agent transactions, updated their terms accordingly, and can point to documented evidence of consent if a dispute lands on their desk.

What Does Your Store Actually Need to Do to Get Agent-Ready?

Preparing for AI-led buying doesn't require a full platform rebuild. It comes down to a handful of practical, sequenced actions:

• Fix discoverability first.

AI agents assess products the way search crawlers do, by reading structured data rather than persuasive copy. Clean, consistent product feeds, accurate schema markup for price, stock, and delivery, and policy pages written in plain language give agents what they need. If an agent can't confidently extract that information from your site, it will complete the comparison elsewhere, and you'll never see that lost sale in your analytics because the customer never technically visited.

• Audit your terms and policies for a hidden assumption.

Most retailers' terms and conditions, refund policies, and dispute processes quietly assume a human is present at the point of purchase to consent in real time. Clauses built around that assumption create exposure the moment an agent is the one clicking confirm. This isn't a rewrite of your entire legal framework; it's a targeted review of the handful of clauses that actually depend on human presence at checkout.

• Stress-test your fraud and authentication rules.

Systems built to flag suspicious human behaviour, unusual location, unfamiliar device, and rapid repeat attempts won't necessarily catch a misconfigured or compromised agent. They may equally block a legitimate one that behaves differently from a typical shopper. Test your fraud rules against agent-shaped traffic before it shows up unannounced in production.

• Get legal, fraud, and payments teams talking to each other.

This kind of purchasing sits across all three functions at once. The businesses that handle a first disputed agent transaction well are the ones where those teams already share a common understanding of what happened and who owns the response.

Frequently Asked Questions

What is agentic commerce and how does it differ from traditional e-commerce? 

Traditional e-commerce still needs a human to browse, decide, and click buy. Agentic commerce hands that job to an AI agent, which researches, compares, and completes the purchase itself under delegated authority. Your site now has to serve machines as well as people.

How can agentic commerce improve the online shopping experience for consumers? Done well, it saves time and cuts decision fatigue. An agent can compare prices, check stock, and apply preferences like budget or delivery speed across dozens of retailers in seconds, something no shopper can realistically do manually before every purchase.

What are the benefits of using AI for managing personal finances and spending? 

AI tools can track spending, flag unusual charges, and enforce budgets automatically, which matters here because agents can be given fixed limits before they're allowed to buy anything. The benefit is control: spending stays within boundaries you set upfront, not after the fact.

How do agentic commerce systems interact with digital assistants and AI bots? 

Agents typically sit downstream of a digital assistant, a voice interface or chat tool that captures the shopper's intent, then hands structured instructions to a separate purchasing agent. Your checkout only ever talks to that purchasing layer, not the assistant itself.

For Magee 1866, the project carried another consideration. The company has a long history and a recognisable identity, so the new website couldn’t feel like a generic eCommerce rebuild with the Magee name added afterwards.

The aim was to modernise the technology behind the store while keeping the experience consistent with the brand customers already knew.

This migration was delivered by the chillicommerce team making an existing Magee 1866 eCommerce operation to Adobe Commerce, creating a single platform that could support retail and wholesale customers across the UK, Ireland and the United States. The work also covered the storefront, checkout, ERP connection, regional store management and several systems used by the wider business.

Why was Migration Needed?

Magee 1866 serves more than one type of buyer.

A retail customer browsing clothing online has different expectations from a wholesale customer placing a business order. Pricing may differ. Account structures are different too, as are ordering habits and the information each customer needs before completing a purchase.

Running these journeys through disconnected systems would create unnecessary work behind the scenes. Product details could become inconsistent, stock figures might fall out of date and teams could end up maintaining the same information in several places.

There was also the question of international selling. Magee needed to manage storefronts for the UK, Ireland and the US without treating every market as a separate eCommerce project.

The new setup therefore needed one central backend, with enough flexibility to handle local currencies, pricing arrangements, account types and customer experiences.

Any rebuild had to improve the experience rather than simply move the existing problems onto a newer platform.

And, naturally, the migration couldn’t discard years of organic search value.

Choosing Adobe Commerce as the Foundation

Adobe Commerce was selected as the main eCommerce platform because it could accommodate the operational complexity behind the Magee business.

The architecture was also planned around future changes. International eCommerce projects rarely stand still after launch. New payment methods appear, customer expectations shift and internal systems change. Building everything too tightly around the original specification would only make the next update harder.

So the Adobe Commerce migration  wasn’t treated as a simple transfer of pages and products. The structure underneath them had to make sense for the way Magee operates now, while leaving room for what might come later.

Rebuilding the Frontend With Hyvä

Adobe Commerce can support complicated stores, but frontend performance depends heavily on how the implementation is built.

For Magee 1866, we used the Hyvä frontend theme alongside Hyvä Checkout. Hyvä replaces much of the heavier frontend approach associated with traditional Magento builds, reducing unnecessary JavaScript and giving developers a cleaner base to work from.

The lighter frontend was especially useful for the product catalogue. Customers need to move between categories, apply filters, view detailed imagery and compare products without the store feeling sluggish.

Hyvä also gives the development team a more manageable frontend for future changes. That matters after launch. A website that performs well on day one but becomes difficult to update six months later isn’t much of an improvement.

Supporting the UK, Ireland and US From One Backend

The new Adobe Commerce setup supports Magee’s UK, Irish and US storefronts through a shared administrative system.

Each market can have its own currency, pricing and local customer experience, while the underlying product and operational information remains centrally controlled. That reduces duplicated work and makes it easier to keep the storefronts aligned.

There’s still space for regional differences. A US customer doesn’t need to feel as though they’ve landed on a UK store with the currency symbol changed, and an Irish customer may require different delivery, payment or content information.

Connecting Adobe Commerce With Fashion Master

The Fashion Master ERP integration was one of the most important parts of the project.

Fashion Master acts as the main operational source for product information, inventory, customer records, orders and pricing. Adobe Commerce needed to exchange that information accurately and quickly enough for the website to reflect what was actually happening inside the business.

Without a reliable ERP connection, even a well-designed storefront can create problems. A product may appear available when there’s no stock left. Prices can differ between internal records and the website. Orders then require manual correction, which slows the team down and increases the chance of mistakes.

Not glamorous, perhaps, but essential. This is the part customers rarely see and businesses notice immediately when it goes wrong.

Bringing Payments, Search and Marketing Into the Platform

Stripe was integrated as the payment system, supporting options including Klarna, PayPal, Apple Pay and Google Pay. Giving customers several recognised ways to pay helps reduce friction at checkout, particularly across different devices and regional markets.

Adobe Live Search supports product discovery within the catalogue, while Fastly handles content delivery and caching. Klaviyo connects the store with Magee’s email and SMS marketing activity, and Reviews.io brings customer reviews into the shopping experience.

Adobe Business Intelligence provides reporting capabilities, with live chat available for customers who need help while browsing or ordering.

These connections weren’t added as isolated website features. Each one had to fit into the wider flow of data, customer activity and order management.

Performance Improvements

The completed platform gives Magee a single Adobe Commerce backend for both retail and wholesale operations.

Its UK, Ireland and US stores can now be managed centrally while retaining their own currencies, customer experiences and pricing requirements. The Fashion Master integration keeps the online store connected with product, stock, order and customer information used elsewhere in the business.

Hyvä and Hyvä Checkout provide a lighter frontend, making it easier for customers to browse the catalogue and complete an order. The surrounding integrations connect payments, reviews, marketing, reporting and product search without turning the website into a collection of disconnected tools.

Most importantly, the migration brought the technology forward without pushing Magee’s identity into the background.

A sizeable technical change, then. From the customer’s side, it should simply feel like a faster and easier way to shop.