Agentic commerce is here. The question is whether you’re designed for it.

What the announcements unlock

Three things came out of NRF 2026 that matter, and they matter in combination.

The Universal Commerce Protocol (UCP). Codeveloped by Shopify and Google, UCP is an open standard that gives AI agents a shared language for commerce. Instead of every retailer building a custom connection to every AI platform, UCP defines how agents discover what a merchant supports and complete a purchase — applying discount codes, loyalty credentials, subscription billing preferences and seller-specific terms like pre-order timing or final sale conditions.

Where a transaction requires customer input — a furniture retailer that needs a delivery window confirmed, for example — UCP provides a standard way for the merchant to specify exactly what’s needed. It accommodates complexity rather than flattening it. Already endorsed by 20+ retailers and platforms, UCP works with any payment processor. This isn’t a Shopify feature. It’s a commerce standard.

Native commerce across every major AI surface. Shopify merchants can now sell natively inside Google AI Mode, Gemini and Microsoft Copilot — managed centrally from a single admin. The checkout doesn’t redirect. The transaction completes inside the conversation. For the first time, the path from “I want this” to “I bought this” can happen entirely within an AI interaction, with your brand logic, pricing rules and post-purchase experience intact.

Agentic commerce for every brand, not just Shopify merchants. Through a new agentic plan, any brand — regardless of platform — can list products in Shopify’s catalog infrastructure and be discoverable and transactable through AI agents. The implication is significant: you don’t need to complete a platform migration to have an agentic presence. When transactions move from human-driven clicks to machine-executed commitments, errors are no longer edge cases — they become training data. This can alter agent behaviour, with weaknesses compounding silently.

You need clean, structured product data and a trust story that agents can read. What these three announcements have in common is that they remove the technical excuses. The protocol exists. The surfaces are live. The infrastructure is available to any brand willing to engage with it seriously. What remains is the harder work — and the harder work isn’t technical.

It’s not only happening in the US. In February 2026, Loblaw — Canada’s largest retailer — launched a PC Express app inside ChatGPT, letting Canadians move from a meal idea to a grocery list to a PC Express cart without leaving the conversation. The following week they announced a second integration, making health, beauty and apparel products shoppable through Google AI Mode and Gemini — the first large Canadian retailer to do so.

Loblaw’s model is worth noting precisely: ChatGPT handles discovery and planning; the transaction completes through PC Express. The brand retains the customer relationship and the data. That architecture — AI as the front door, your infrastructure as the back end — is a deliberate choice, not a limitation. And it’s available to any retailer willing to build for it.

The temptation for most organizations is to treat agentic readiness as an optimization project — a new channel requiring a new checklist. Update the schema. Publish an llms.txt file. Make sure the crawlers can access your content. Done.

Some of that work matters. But the organizations that approach this as an extension of SEO are going to find themselves ready for a game that’s no longer being played.

Where it overlaps

The SEO disciplines that built content authority still apply — but the reason they apply has changed.

• Structured data and schema markup remain essential, not because they help rankings, but because they’re how agents parse your catalog. If your product attributes are incomplete or inconsistent, agents can’t represent you accurately.
• Content quality matters more than ever. But the measure isn’t keyword density — it’s whether an LLM can extract a clear, defensible answer from your content when someone asks a direct question about your category.
• Third-party authority signals — reviews, citations, expert endorsements — carry weight because agents are built on retrieval-augmented generation. They synthesize from what’s been written about you across the web, not just what you’ve written about yourself.
• Bing Webmaster Tools verification is the most underrated quick win. No Bing index means you’re invisible to ChatGPT and Perplexity. Most organizations haven’t touched it in years.

Where it breaks completely

Here is where the SEO analogy falls apart — and where treating this as an optimization project becomes actively harmful.

Intent isn’t a keyword. In search, you optimize for the words people type. In agentic commerce, you’re optimizing for what people mean — the underlying job to be done behind a request. An agent asked to “find me a sustainable yoga mat under $50” isn’t running a keyword match. It’s reasoning across product attributes, brand claims, third-party verification and user preferences simultaneously. Your product title is table stakes. Your values story is the differentiator.

Agents don’t browse. The traditional commerce funnel — awareness, consideration, purchase — assumed a human moving through stages over time. Agents compress or entirely eliminate stages. In traditional commerce, broken experiences surface as recoverable. For example, abandoned carts or refunds. In agentic commerce, by the time a transaction surfaces to a user for approval, the consideration has already happened inside the model. The implication: the content and signals that used to guide a human through a funnel now need to exist in machine-readable form before the conversation even starts.

You can’t buy your way in. Search advertising gave brands a mechanism to appear regardless of organic merit. Agentic commerce, at least in its current form, shifts that dynamic significantly. As Harley Finkelstein put it: “[Agentic commerce] is not necessarily based on who is the largest company. It’s based on what consumers are looking for.” ¹

Paid and organic channels don’t disappear — they remain important for driving awareness and top-of-funnel reach. But inside agentic experiences, recommendation is increasingly driven by fit, trust and values alignment rather than spend alone. That creates an opening for brands with a genuine story — and a new kind of pressure on brands whose visibility has relied primarily on paid scale.

The SEO → AEO → ACO progression is real and sequential. Search engine optimization got you ranked. Answer engine optimization gets you cited in LLM responses. Agentic commerce optimization gets you transacted through AI marketplaces. Each layer builds on the last — but the KPIs are completely different. ACO isn’t measured in impressions or CTR. It’s measured in share of recommendation, agent-driven conversions and trust scores. Most organizations don’t have visibility into any of these yet.

There is a principle at the heart of good experience design: every screen is a conversation between an application and its user. The best designers don’t wait for users to articulate what they need — they anticipate intent, remove friction and shape that conversation so the right answer arrives before the user knows to ask for it. The brands that internalized this built loyalty. The ones that treated digital as a broadcast medium lost customers to whoever made the experience feel effortless.

Agentic commerce is that same conversation — except the user is now an AI acting on someone’s behalf. And this changes the design brief in a fundamental way.

An agent isn’t a neutral search engine. It’s a proxy for the values, preferences and constraints of the person it represents. When someone delegates a purchase to an agent — explicitly or through learned behaviour — they’re encoding a set of priorities: sustainability over speed, ethics over price, community impact over convenience. The agent surfaces those priorities in every transaction it runs. Which means a brand’s ability to be chosen depends not just on whether it can be found and trusted, but on whether its values are legible to a machine making decisions on someone else’s behalf.

That’s not an optimization problem. That’s an architectural one.

Three filters. One design brief.

Think of it this way. Every agent runs three filters on every brand, in sequence, before it recommends a purchase. Not as a formal checklist — but as the emergent logic of how these systems evaluate fit. Getting any one wrong means not clearing the next.

Filter 1 Can I find you?

The entry fee — necessary but not sufficient

Your product data needs to be structured for agent consumption, not human browsing. UCP-compliant catalog data, complete schema markup, real-time inventory integrity, AI crawler access, answer-first content architecture. This is the work most organizations are focused on — and it matters. But it’s a commodity within 18 months. Every serious competitor will have it. Passing Filter 1 qualifies you to compete. It doesn’t win you anything.

Filter 2 Can I trust you?

Where revenue is protected or lost

When an agent promises a customer a product at a price with a delivery window, your operational infrastructure has to back that promise. Trust is not only about reliability. When agents take actions on your behalf, considerations like identity assurance, transaction integrity, fraud controls and continuous monitoring are also part of it. Broken promises at agentic scale aren’t support tickets — they’re the signal that trains models away from recommending you.

This layer requires real-time inventory accuracy, structured fulfillment data, machine-readable return and dispute policies, and identity/loyalty integration that lets agents transact on behalf of known customers. Most organizations haven’t achieved this coherence even in non-agentic contexts. It’s the trust architecture that determines whether agentic commerce helps or hurts you.

Filter 3 Should I choose you?

The only sustainable moat

This is where the design challenge lands hardest — and where most organizations have done the least work. Filter 3 isn’t about discoverability or reliability. It’s about whether your brand has a values story that an agent can read, verify and match against the preferences of the person it’s buying for. Weakly governed values claims don’t just risk reputational damage – there is risk that your agents are trained contrary to your brand.

What does your brand optimize for? Sustainability, fair wages, community impact, ethical sourcing? The answer matters — but only if it exists as structured, verifiable, machine-readable data. Third-party certifications. Schema-encoded claims. Verified review signals with timestamps. This isn’t a marketing exercise. It’s an architectural decision about what your brand stands for and how you prove it to a non-human evaluator.

Fewer than 1% of organizations are competing at this layer today. Citation patterns in AI systems self-reinforce over time — the brands that establish values-based authority now will be disproportionately surfaced as agent populations grow. That window doesn’t stay open.

The Organizational Reality

What this requires

The three filters map to three different organizational muscles — and most organizations are only exercising one of them.

Filter 1 is a technology and content operations challenge. Most organizations have the teams for it. It’s a matter of priority and sequencing.

Filter 2 is a cross-functional operations challenge. It requires alignment between commerce technology, supply chain, finance and customer service that most organizations haven’t achieved even without agentic commerce in the picture. The urgency is new. The organizational gap isn’t.

Filter 3 is a leadership challenge. It requires decisions about what the organization genuinely stands for — decisions that tend to get deferred because they’re uncomfortable — translated into verifiable, structured claims that exist outside of a brand guidelines document. This is not a task that can be delegated to a content team or an agency. It requires the organization to have an answer to the question an agent will ask: “why should I choose you over a functionally equivalent competitor?”

The organizations that win agentic commerce will be the ones that pass all three filters coherently — where their values inform their trust architecture and controls are embedded in design, which then informs implementation. Where the whole stack tells the same story. That’s not a technology deployment. That’s a transformation.

Contributors:
Pramod Gopalakrishna - Director, Blockchain and AI transformation, Technology Consulting
Kyle Morton - Staff Consultant, Risk Consulting

The Bottom Line

The front door has changed. Have you?

The infrastructure is here. UCP is live. The major AI surfaces are open for commerce. Since January 2025, AI-driven traffic to Shopify stores is up 6x and orders are up 11x — and that’s before most brands have done anything deliberate to capture it.

The organizations that move now — not just on the technical foundations but on the trust architecture and values positioning above them — will establish citation patterns and agent preference that compound over time. The ones that treat this as an SEO refresh will be well optimized for a conversation that’s already moved on.

Three filters. Can I find you? Can I trust you? Should I choose you? Answer all three, in order, with evidence an agent can genuinely read. That’s the design brief.