What is agentic commerce?

Agentic commerce is the shift from consumers manually searching and browsing to AI agents — such as Google Gemini, Microsoft Copilot, and ChatGPT — discovering, evaluating, and completing purchases on a user's behalf, often without the customer ever visiting a brand's website.

What is the Universal Commerce Protocol (UCP)?

UCP is an open standard co-developed by Google and Shopify, launched in January 2026, that defines how AI agents interact with commerce systems — discovering products, building carts, processing payments, and managing orders. It is supported by a coalition including Walmart, Target, Wayfair, and Etsy, and is compatible with REST, MCP, and Agent2Agent frameworks. Merchants onboard through Google Merchant Center.

What is the difference between UCP and ACP?

UCP (Universal Commerce Protocol), led by Google and Shopify, is designed for breadth and interoperability — one merchant endpoint compatible with any adopting platform. ACP (Agentic Commerce Protocol), developed by OpenAI and Stripe, is optimised for depth and streamlined in-agent transactions, powering ChatGPT Instant Checkout across over a million Shopify merchants. Both preserve the merchant as seller of record, and both are increasingly interoperable — Microsoft Copilot uses elements of both simultaneously.

What are Microsoft Copilot Brand Agents?

Brand Agents are AI-powered shopping assistants trained on a retailer's product catalogue and embedded directly into their website or commerce experience. They answer customer questions conversationally — covering fit, availability, pricing, and variants — and guide shoppers toward purchase. Early data from Microsoft suggests Brand Agent-assisted sessions generate around a 2x lift in conversion rates compared to unassisted sessions.

What is Identity Linking and why does it matter?

Identity Linking is a feature within UCP that allows consumers to connect their existing loyalty accounts — via OAuth 2.0 authentication — to agent-mediated transactions. This means loyalty discounts apply automatically, points accrue, and the purchase creates a verified record in the retailer's CRM. Without Identity Linking, agent-completed purchases may bypass a brand's owned customer data infrastructure entirely. Microsoft demonstrated this with Target Circle in Copilot in early 2026.

How long does a UCP integration take?

For Google AI Mode specifically, the foundational requirement is three core API endpoints. Teams with clear ownership and executive support are typically moving from planning to testing in approximately four weeks. Shopify merchants can achieve basic protocol eligibility in days to weeks at relatively low cost. Custom enterprise stacks are a 3-to-6 month project. The recommended approach is phased: clean product data first, then real-time APIs, then checkout, then identity linking.

Do I need to replace my payment system to support agentic checkout?

No. When a consumer pays via Google Pay or a similar interface during an agent-assisted checkout, merchants are not required to rebuild their payment infrastructure. A tokenised payment credential is generated and passed through the merchant's existing payment service provider. One important step: merchants must register as a Google Pay merchant, which can take up to 30 days, so this should be initiated at the start of any integration project.

Is agentic commerce available in the UK and Europe?

As of mid-2026, most fully operational agentic checkout systems — including UCP's Google AI Mode integration and ACP's ChatGPT Instant Checkout — are primarily live in the United States. UK and EU rollout timelines have not been confirmed. Regulatory factors including the EU AI Act, GDPR Article 22, and PSD2 Strong Customer Authentication requirements are shaping how in-agent payment flows will need to be designed for European markets. Brands are advised to build foundational infrastructure — data quality, APIs, identity systems — now, in preparation for eventual rollout.

What types of purchases are best suited to agentic commerce?

Agentic commerce is best suited to routine, goal-oriented purchases: replenishment orders, commodity products, known-item buying, and repeat transactions where convenience matters more than discovery. High-consideration or emotionally driven purchases — luxury goods, furniture, complex electronics — are less likely to move to fully autonomous agent-mediated flows in the near term, as consumers in those categories typically want exploration and reassurance.

Agents at the Checkout: What Agentic Commerce Means for Your Brand in 2026

Bradley Slinger
May 13
8 min read

In a previous article, I made the case that the web was never really designed for AI agents. Most digital infrastructure, JavaScript-heavy front ends, rendered pages, fragmented authentication flows, was built for humans navigating visually, not machines processing information programmatically at scale.

But that conversation only covered discovery.

What happens when those same agents stop browsing and start buying?

That shift is already underway. Commerce is no longer something customers necessarily "visit." Increasingly, it is becoming something embedded directly into conversations, handled by AI systems acting on a user's behalf, often without the customer ever landing on a brand's website at all.

For brands heading into 2026, the important question is no longer whether agentic commerce is real. It is whether your business is visible inside it.

The Shift Happened Quietly

There was no single launch moment. Instead, agentic commerce arrived gradually, a protocol announcement here, an embedded checkout flow there, until the underlying trend became difficult to ignore.

693% increase in AI-driven e-commerce traffic during the 2025 holiday season - Adobe, analysing over a trillion retail site visits. 8,000% YoY growth in agent-assisted browser sessions - Microsoft telemetry. 7,851% increase in automated commerce activity - HUMAN Security, independent analysis.

The percentages are dramatic, and context matters. These systems are still growing from a relatively small base. But the direction is unmistakable.

One of the fastest-growing "buyer groups" online is no longer a demographic. It is software.

And software behaves very differently from people.

AI agents do not browse aimlessly or react to promotional banners. They evaluate structured information, compare options against defined criteria, and execute decisions. If your product data is inconsistent or inaccessible to the protocols these agents rely on, you are not simply ranked lower. In many cases, you are excluded entirely.

Commerce Needs a Shared Language

Two protocols are beginning to define how AI agents interact with commerce systems.

	UCP	ACP
Led by	Google + Shopify coalition	OpenAI + Stripe
Launched	January 2026, NRF	September 2025
Philosophy	Breadth and interoperability	Depth and payment integration
Powers	AI Mode in Search, Gemini	ChatGPT Instant Checkout
Strength	One endpoint, any platform	Stripe-native, 1M+ Shopify merchants

They are often framed as competitors, but in practice they look more like different dialects of the same language. Both define how agents discover products, build carts, authenticate transactions, and confirm orders. Both preserve the merchant as the official seller of record, brands still own fulfilment, customer relationships, and transaction data.

Interoperability is already emerging. Microsoft Copilot, for example, uses UCP-compatible product discovery while leveraging ACP-style checkout flows underneath. Stripe sits beneath both as a shared payment rail that neither protocol owns.

The question for brands is not which protocol wins. It is whether your product data is structured well enough to be readable by either of them.

Microsoft's Brand Agents Make the Shift Tangible

Protocols are abstract. Brand Agents are easier to visualise.

Initially launched for Shopify merchants and expanding toward WooCommerce, Brand Agents are AI-powered shopping assistants embedded directly into retailer experiences. Instead of a customer navigating filters and product pages, the agent answers conversationally:

Does this fit true to size?
Is the fabric lightweight?
What colours are available in my size?
Is this in stock for next-day delivery?

Responses are grounded in the retailer's actual catalogue — not generic AI assumptions.

Early conversion data from Microsoft:

Brand Agent-assisted sessions generated around a 2x increase in conversion rates on average
Retailer Alexander Del Rossa reported a 3x uplift in Brand Agent-assisted journeys
Copilot-assisted sessions with high purchase intent were significantly more likely to convert

Treat those figures carefully as they come from early deployments and platform-owned reporting, which naturally skews optimistic. But even if the eventual impact settles well below those numbers, the direction of travel is clear: conversational commerce is moving from experimentation into measurable commercial performance.

The Bigger Strategic Issue: Identity

Most conversations about agentic commerce focus on checkout friction. The more important long-term issue is identity.

Here is the risk. If an AI agent completes a purchase entirely on a customer's behalf, brands can lose direct access to the relationship — no email capture, no loyalty association, no CRM enrichment, no first-party signal. At scale, that is a serious erosion of owned customer data built over years.

This is where UCP's Identity Linking becomes strategically important.

Using OAuth-based authentication flows, customers can connect existing loyalty accounts directly to agent-driven transactions. A linked customer automatically applies discounts, accrues reward points, and maintains continuity inside the retailer's CRM.

Microsoft demonstrated this with Target in early 2026. A Copilot user could link their Target Circle account mid-conversation without breaking the shopping flow. Target received a verified, loyalty-attributed transaction rather than an anonymous guest checkout.

That reframes the entire proposition. Agent-mediated commerce does not have to mean disintermediated commerce. Brands that invest in identity infrastructure may not lose the customer relationship — they simply access it through a different interface layer.

What Agent-Readiness Actually Looks Like

Agent-readiness is essentially a data and infrastructure problem with three layers.

Layer 1: Product data quality Agents rely on clean SKUs, GTINs, consistent variant structures, normalised attributes, and accurate inventory states. If those systems are fragmented internally, agents struggle to interpret your catalogue confidently.
Layer 2: API performance Real-time catalogue queries, cart actions, and checkout requests all require stable, low-latency infrastructure. Checkout logic must be idempotent — meaning duplicate agent requests do not create duplicate orders.
Layer 3: Identity and loyalty infrastructure OAuth flows, authorisation metadata, account-linking experiences, and CRM integration. This is the layer most brands build last, which is the right sequencing.

Implementation varies significantly by starting point:

Merchant type	Timeline	Indicative cost
Shopify / platform-native	Days to week	£5k - £25k
Middleware / adapter approach	4 - 8 weeks	£30k - £120k
Custom enterprise stack	3 - 6 months	£120k - £400k+

The smarter approach is phased: clean the data first, expose real-time pricing and inventory next, enable checkout compatibility, then layer identity linking. The businesses making fastest progress are not necessarily the most technically advanced — they are the ones that aligned internal ownership early.

The Practical Starting Point

One insight emerging consistently from merchants already working with UCP is how manageable the first step actually is.

For Google AI Mode integrations specifically, the foundational requirement is three core API endpoints. Teams with clear ownership and executive support are moving from planning to testing in roughly four weeks.

The organisational side often matters more than the technical side. Three roles consistently determine pace:

An executive sponsor who can unblock legal and compliance questions quickly
A project manager coordinating engineering, product, and legal workstreams
An e-commerce lead who understands existing Merchant Center infrastructure

On payments: consumers may see Google Pay during agent-assisted checkout, but merchants are not rebuilding payment systems from scratch. Tokenised credentials pass through existing PSP infrastructure. The one action that consistently catches teams off guard, payment credential registration, can take up to thirty days. Start it at the beginning of the project.

And importantly, UCP integrations are not platform-exclusive. The argument for building is not "integrate with Google." It is "build once for any compatible ecosystem" — including Microsoft, Meta, and Amazon, all of which have now joined the UCP Tech Council.

Europe Will Move More Carefully

For UK and EU brands, geography matters. Most fully operational agentic checkout systems are currently concentrated in the US. The delay in Europe is partly regulatory:

The EU AI Act introduces stricter requirements around automated decision-making and transparency
GDPR Article 22 creates complications around fully autonomous transactional systems
PSD2 / Strong Customer Authentication adds friction to in-agent payment flows

That does not mean agentic commerce bypasses Europe. It means the implementation model will evolve differently, more consent layers, more explicit confirmations, more human oversight inside the flow.

For European brands, the practical takeaway is straightforward: build the foundational infrastructure now, while recognising that rollout timelines may lag the US. The underlying investments improve digital performance across every channel regardless.

The Sceptics Have a Point

Not everyone believes this represents a major behavioural shift, and that scepticism is worth engaging with honestly.

Agentic commerce is unlikely to dominate emotionally driven or high-consideration purchases anytime soon, luxury, premium fashion, complex electronics, furniture. People still want exploration and emotional engagement in those categories.

The stronger early use cases are routine and goal-oriented: replenishment, commodity products, known-item purchases, repeat orders, utility-focused shopping where convenience outweighs discovery.

That is a large and commercially meaningful subset of retail. And in those scenarios, the implications for visibility are the same.

The New Visibility Problem

Every major digital channel has historically created a new form of visibility competition.

Retail fought for shelf space. Search fought for rankings. Social commerce fought for feed presence.

Agentic commerce introduces a different dynamic. In previous channels, poor optimisation meant reduced visibility — lower rankings, smaller reach. You were still present.

In agentic systems, poor infrastructure can mean total absence. An AI agent does not "half understand" broken product data. It ignores it completely and moves on to a competitor whose data it can trust.

That is why the conversation matters now, even while the market is still early. The infrastructure investments required for agent-readiness — structured data, real-time APIs, identity systems, protocol-compatible commerce flows — are becoming foundational digital capabilities regardless of whether agentic commerce becomes the dominant channel in 2026 or 2030.

The long-term question is not whether AI agents will participate in commerce.

They already are.

The real question is whether your brand is structurally prepared to be included when they do.

ARRAY Analytics helps brands and marketing teams navigate the shift toward AI-driven search, discovery, and commerce, from GEO strategy and LLM visibility tracking to structured data audits and agent-readiness assessments.

Appendix: Key terms and definitions

Agentic Commerce: A paradigm of e-commerce where intelligent AI agents autonomously handle product discovery, evaluation, negotiation, and purchase execution.
UCP: An open standard co-developed by Google and Shopify to facilitate the entire agent-to-merchant commerce lifecycle.
ACP: An open standard led by OpenAI and Stripe, optimized for streamlined, in-agent checkout experiences.
Brand Agents: AI-powered shopping assistants trained on a brand's catalog and deployed on the merchant's website.
Copilot Checkout: A feature enabling users to complete purchases directly within the Microsoft Copilot chat interface.
Identity Linking: A standardized mechanism (typically OAuth 2.0) connecting a user's platform identity with their merchant account.
Merchant-of-Record (MoR): The legal entity responsible for a customer transaction.
Delegated Payments: A model where an agent tokenizes payment credentials into a secure vault token passed to the merchant.
Payment Handler: A UCP concept defining how a payment instrument is processed.
Amazon "Buy for Me": A feature where Amazon's AI agent makes a purchase on a consumer's behalf from a third-party brand's website.