Agentic Commerce: Full Commerce Guide

TL;DR — Key takeaways

AI-referred traffic to US retail sites grew 805% year-over-year on Black Friday 2025 — agentic commerce is not a forecast, it is operational today.
Two protocols are live: OpenAI + Stripe’s ACP (September 2025) and Google’s UCP (January 2026) — every brand needs a position on both.
The real risk is not which protocol wins. It is whether your product data is structured well enough for any agent to find, evaluate, and transact with your brand.
The Full Commerce Stack (Signal → Decision → Transaction) gives brand managers a practical architecture for agent readiness without a full technical overhaul.
McKinsey estimates the global agentic commerce opportunity at $3–5 trillion by 2030 — the brands building agent infrastructure now will capture a disproportionate share.

A brand manager at a mid-sized cosmetics company told me something that stopped me mid-sentence last quarter. Her products were priced competitively, her reviews were solid, and her Amazon listing looked polished. Yet when she asked ChatGPT to recommend a vitamin C serum in her category, her product did not appear. Not on page one. Not on page two. Not at all. Her competitor — smaller catalogue, simpler packaging — showed up first. The difference had nothing to do with quality. It had everything to do with data structure.

That gap is what agentic commerce actually means for brand managers. Not the consumer-facing spectacle of AI agents clicking buy now on someone’s behalf. The real transformation is happening on the sell side, and most brands are not seeing it yet.

Why “AI Agents Buy Stuff” Is the Wrong Frame

Nearly every article on agentic commerce opens with the same scene: a consumer delegates shopping to an AI, the agent compares options, negotiates, and completes the purchase. The narrative is real. Morgan Stanley projects that nearly half of online shoppers will use AI shopping agents by 2030. ChatGPT’s Instant Checkout has been processing transactions for 900 million weekly users since September 2025. During Cyber Week 2025, 20% of global orders were influenced by AI agents, according to eMarketer.

But here is what gets left out: agents do not find products the way search engines do. They do not rank by keyword density or backlink count. They read structured data. They query APIs. They evaluate attribute completeness, trust signals, and transactional readiness in milliseconds — and they skip anything that cannot answer their questions cleanly.

What surprises me, even now, is how many brand leaders are watching the protocol wars between OpenAI’s ACP and Google’s UCP with the wrong question. They ask: which protocol should we bet on? The right question is: can any agent read our catalogue at all?

The Full Commerce Stack: A Framework for Agent-Ready Brands

At Epinium, we have been helping brands and manufacturers restructure for AI-driven distribution since before the current wave of consumer-facing agents. What we converged on is a three-layer architecture we call the Full Commerce Stack.

Layer 1 — Signal Layer. This is the foundation: structured product data, rich and complete attributes, AI-readable pricing signals, and clean taxonomy. If your product has 40 attributes and only 12 are filled, you are invisible to structured queries. Agents evaluating thousands of SKUs in a category do not guess at missing values — they exclude them.

Layer 2 — Decision Layer. Once an agent can read your product, it needs to evaluate it. This layer covers AI-optimised content (explicitly comparative, not keyword-stuffed), review signal architecture, and competitive advantage encoding — the articulation of why your product wins in specific use cases. Vague copy kills here. An agent comparing five vitamin C serums needs to know: does this contain L-ascorbic acid above 15%? Is it fragrance-free? What is the dermatologist approval rate? If your listing does not say it plainly, the agent will not infer it.

Layer 3 — Transaction Layer. This is where most brands have the biggest gap in 2026. Agents completing purchases need instant inventory confirmation, reliable fulfilment signalling, and in some architectures, API-accessible checkout hooks. A brand with 24-hour inventory update latency is a liability in agentic commerce — the agent will route to a competitor who can confirm stock in real time.

805%

growth in AI-referred traffic to US retail sites — Black Friday 2025 vs. 2024

Source: eMarketer 2025

Agentic Commerce in 2025–2026: What Actually Changed

ACP Goes Live — September 2025

OpenAI and Stripe launched the Agentic Commerce Protocol in September 2025, embedding it directly into ChatGPT for 900 million weekly users. ACP enables ChatGPT to complete purchases, check inventory, and process transactions without leaving the conversational interface. Brands on Shopify and select partner platforms were first movers. If your brand is not on an ACP-integrated platform, ChatGPT’s agent cannot transact with you.

Google’s UCP Coalition — January 2026

Google announced the Universal Commerce Protocol in January 2026, backed by a coalition of retailers and platform operators. UCP works across Google Search AI Mode, Gemini, and third-party agents. Unlike ACP, it does not require brands to be on a specific platform — but it does require structured data compliance and API accessibility. Early partners include major FMCG manufacturers and Amazon aggregators.

Amazon Rufus and the Quiet Readiness Signal

Amazon has not launched a public agentic checkout protocol, but Rufus has been surfacing products based on attribute completeness and review quality — not traditional ad spend — since early 2025. Brands that invested in A+ Content 2.0 and Virtual Bundles are seeing disproportionate Rufus visibility. What Amazon is doing quietly, open agents will do loudly by Q4 2026.

AI Commerce Revenue Is Already Quadrupling

AI platform ecommerce sales reached $20.57 billion in 2026 — nearly quadruple 2025 figures. This is not a rounding error. It is a structural shift in where purchase intent originates. Brands tracking analytics will have seen this in referral data already.

Traditional Commerce vs. Agentic Commerce: What Shifts for Brands

Dimension	Traditional Commerce	Agentic Commerce
Discovery	Keyword search, paid ads, influencer	Structured attribute queries, API calls
Evaluation speed	Human browsing (seconds to minutes)	Millisecond data parsing at scale
Key content asset	Attractive images, emotional copy	Complete structured attributes, explicit comparative claims
Trust signals	Star ratings, brand familiarity	Verified specs, structured review data, return rate signals
Transaction trigger	Human checkout decision	Agent API call via ACP or UCP protocol
Inventory requirement	24-hour update cycles tolerable	Real-time confirmation required

Epinium data

In Q1 2026, across brand portfolios managed via Epinium Platform, catalogs with structured attribute completeness above 85% received 3.2× more AI-referred sessions than those below that threshold. The effect was most pronounced in personal care, home goods, and nutritional supplements — categories where agent queries are highly attribute-specific.

FREE DIAGNOSIS

Is Your Brand Invisible to AI Agents?

In 30 minutes, the Transform team audits your Full Commerce Stack readiness and shows exactly where agents are skipping your products — and what to fix first.

How Transform works → ✓ 30 min ✓ No cost ✓ Dedicated AI Director

The Myth That Is Costing Brands the Most Right Now

Here is where most brands get it wrong. They treat agentic commerce as a channel optimisation problem — something to hand to the SEO team or the Amazon agency. It is not. It is a data architecture problem sitting at the intersection of operations, marketing, and technology.

Being on Amazon does not make you agentic-commerce ready. Amazon’s Rufus is one agent in a closed ecosystem. ACP and UCP agents access your products wherever they live — your DTC site, your retail media presence, your product information management system. A polished Amazon listing does not compensate for broken product feeds on your own domain. The brands winning in agentic commerce are not the ones with the biggest ad budgets. They are the ones with the most complete, machine-readable product intelligence.

In a project with a cosmetics brand, we restructured their product data architecture — not adding new content, just making existing information machine-readable and attribute-complete. Appearance in AI-generated recommendation outputs increased 4× within two months. The products did not change. The data did.

What we see at Epinium is a consistent pattern: brands that invested in catalogue infrastructure before the agent wave (2023–2024) are now seeing compounding returns. Brands scrambling to retrofit in Q2 2026 are looking at 4–6 months of structured work before agents can reliably find and evaluate their products.

For deeper strategic context, see our Agentic Commerce Strategy playbook and the guide on how brands get chosen by AI agents.

Frequently Asked Questions About Agentic Commerce

What is agentic commerce?

Agentic commerce refers to commercial transactions that are discovered, evaluated, and completed — fully or partially — by AI agents acting on behalf of users or businesses. The agent (a system like ChatGPT, Gemini, or a custom enterprise AI) is the buyer-side intermediary. It queries product data, evaluates fit, and initiates purchase through a protocol like ACP or UCP. For brands, this means products need to be machine-readable and transactable via API, not just visually appealing on a product page. Agentic commerce is not a new storefront. It is a new buyer.

How is agentic commerce different from traditional ecommerce?

Traditional ecommerce optimises for human perception — attractive images, emotional copy, intuitive UX. Agentic commerce optimises for machine comprehension — complete structured data, explicit attribute signals, real-time inventory APIs. A product can look perfect to a human buyer and be completely invisible to an AI agent. The underlying sales channels are the same; the language you need to speak to be evaluated is entirely different.

Which AI agents are already completing purchases today?

ChatGPT Instant Checkout via ACP and Stripe has been live since September 2025, serving 900 million weekly users. Shopify merchants integrated into OpenAI’s ecosystem were first movers. Google Gemini processes commercial queries via UCP with select retail partners since January 2026. Amazon Rufus influences purchase decisions at scale within Amazon’s own marketplace. Perplexity’s shopping feature and enterprise procurement AI tools are also completing B2B transactions via structured product data.

ACP or UCP — which should my brand prioritise?

Both, eventually. ACP (OpenAI + Stripe) has the larger consumer-facing deployment today at 900 million ChatGPT users. UCP (Google) has broader protocol ambitions and more enterprise applicability. Both protocols read the same underlying asset: well-structured product data with complete attributes and accessible inventory APIs. Build the data foundation first. Protocol compatibility follows from that.

How do I know if my products are agent-discoverable right now?

The fastest diagnostic: open ChatGPT, describe your product category with specific attributes — “fragrance-free vitamin C serum above 15% L-ascorbic acid under £30” — and see if your brand surfaces. Run the same query in Perplexity and Gemini. If you do not appear in any of the three, your structured data is likely incomplete. A more systematic audit covers attribute completeness in your PIM, schema markup on your DTC site, and Amazon backend search term fields.

If we already sell on Amazon, are we prepared for agentic commerce?

Not necessarily. Amazon Rufus operates within a closed ecosystem. ACP and UCP agents access products across the open web — including your DTC site, Google Shopping feed, and third-party retail partners. A brand fully optimised for Amazon but with a thin DTC presence and weak structured data elsewhere will be invisible to the majority of agentic commerce traffic. Amazon readiness is necessary but not sufficient.

What is the minimum viable first step toward agentic commerce readiness?

Complete your product attribute data — every field, every SKU. Agents filter by completeness. A product with 12 of 40 attributes populated will lose to a weaker product scoring 38 of 40, every time. The second highest-leverage action: implement JSON-LD Product schema (including Offer and AggregateRating) on every product page of your DTC site. That single change makes your products readable by every major AI crawl and agent query system currently operating.

What is the Full Commerce Stack and how long does it take to implement?

The Full Commerce Stack is Epinium’s three-layer architecture for agent readiness: Signal (structured product data and AI-readable catalogues), Decision (agent-optimised content and competitive positioning), and Transaction (real-time inventory and API-accessible checkout). For a mid-sized brand managing 500 to 2,000 active SKUs, Signal and Decision layers typically take 10 to 14 weeks to implement, depending on current data fragmentation. Transaction layer readiness depends on existing ERP and fulfilment architecture.

What happens to brands that do nothing about agentic commerce in 2026?

AI platform ecommerce sales reached $20.57 billion in the US alone in 2026 — nearly 4× the 2025 figure. McKinsey projects the global agentic commerce opportunity will exceed $3 trillion by 2030. Brands excluded from agent discovery in 2026 are not just missing short-term revenue — they are allowing competitors to build agent-preference data that compounds over time. Agents learn which brands they can transact with reliably. Not adapting is not a neutral decision.

Can smaller brands compete in agentic commerce without a large tech team?

Yes — and this is one of the more encouraging structural aspects of the shift. Agentic commerce is primarily a data quality problem, not an engineering problem. A small brand with disciplined PIM strategy, complete attribute data, and well-structured schema markup can outperform a much larger brand with better visual assets but fragmented data infrastructure. The barrier is attention and execution, not budget. Smaller brands that move fast on data quality in the next six months will be meaningfully ahead of larger competitors still debating internal ownership.

The brands that capture the most from agentic commerce will not be those with the most sophisticated AI strategies. They will be the ones that made the unglamorous infrastructure investments early — complete data, readable catalogues, accessible APIs — and positioned themselves as the easiest option for agents to work with. That window is 2026. It will not stay open indefinitely.

TRANSFORM BY EPINIUM

Find Out Where AI Agents Are Skipping Your Brand

Brands in Epinium’s portfolio with Full Commerce Stack implementations saw 3.2× more AI-referred sessions in Q1 2026. Book your free 30-minute diagnosis.

Book free diagnosis →

30 min · No cost · Personalised diagnosis