Agentic commerce 2027: what changes for online sellers

TL;DR — Agentic commerce is a purchase carried out by an AI agent on the shopper's behalf: discovery, comparison, add-to-cart and in some cases payment happen inside ChatGPT, Gemini or a proprietary assistant. Two protocols are competing to become the standard: ACP (OpenAI + Stripe) and UCP (Shopify/Google). For a merchant the priority is not "adopting AI": it is making catalogue, feeds and checkout readable and callable via API. Armah prepares Magento/Adobe Commerce and Hyvä stores for this scenario with a focused technical audit.

Free technical audit · Talk to an engineer

What agentic commerce is (and what it is not)

It is not a chatbot on your site. It is an interface shift: traffic no longer arrives only from a SERP and a click, but from a software agent that queries data sources, evaluates alternatives and completes the transaction. The store stops being only a page and becomes an endpoint.

Three maturity levels:

1. Visible — the agent can read and cite your products (feeds, structured data, extractable content).
2. Queryable — the agent can verify availability, price and variants in real time via API.
3. Transactional — the agent can create an order and pay through a shared protocol (ACP/UCP).

Most mid-market merchants today do not get past level 1.

ACP vs UCP: which agentic commerce protocol matters

	ACP (Agentic Commerce Protocol)	UCP (Universal Commerce Protocol)
Backers	OpenAI, Stripe	Shopify, Google/Gemini ecosystem
Surface	Instant Checkout in ChatGPT	AI Mode / Google assistants
Focus	delegated payment and order creation	universal description of catalogue and commerce
Impact on you	product/order endpoints + delegated payment handling	far richer feeds and product schema

You do not need to bet on one of them. Both require the same foundations: clean product data, reliable availability, documented cart and order APIs, machine-readable shipping and returns policies.

What changes for a Magento and Adobe Commerce store

• Product feed: from "good enough for Google Shopping" to complete — GTIN, variants, materials, dimensions, delivery times, returns.
• Structured data: Product, Offer, AggregateRating, ShippingDetails, MerchantReturnPolicy consistent with the feed. Mismatches get the offer discarded.
• Real-time availability: wrong stock for an agent is a cancelled order, not an abandoned cart.
• APIs: stable, versioned search, price, cart and order endpoints with sensible rate limits.
• Checkout: the ability to accept an order created "from outside", with delegated payment and adequate fraud checks.
• Content: extractable product pages and guides with short answers at the top. Same work as AEO.

90-day roadmap to become agent-ready

Days 1-30 — Diagnosis Audit of feed and structured data, check how answer engines read the site today, map existing APIs, review AI crawler access in robots.txt.

Days 31-60 — Foundations Clean up catalogue attributes and PIM, complete Product/Offer schema, reliable stock sync, answer-first pages for commercial queries.

Days 61-90 — Enablement Documented catalogue and order endpoints, sandbox for agentic testing, tracking of traffic coming from AI assistants, ACP/UCP decision based on your channel mix.

What it costs to make your store agent-ready

Workstream	Indicative range
Agent-readiness audit (feed, schema, API)	€ 2,500 – € 6,000
Catalogue and product data clean-up	€ 5,000 – € 25,000
Structured data + AEO content	€ 4,000 – € 15,000
Commerce APIs and agentic checkout integration	€ 15,000 – € 80,000

Ranges depend on SKU count, PIM quality and the state of ERP integrations.

Mistakes to avoid in agentic commerce projects

• Blocking AI crawlers "to be safe" and disappearing from answers.
• Optimising content while leaving feeds and stock dirty: the agent stops at the data.
• Building an internal chatbot and believing you have done agentic commerce.
• Waiting for "the winning standard": the groundwork is identical for ACP and UCP.

Agentic commerce FAQ

Is agentic commerce already worth money today? Transaction volumes are still small, but the preparatory work (data, feeds, schema, APIs) immediately improves Google Shopping, SEO and conversion. It is a double-return investment.

Do I have to choose between ACP and UCP? Not now. Build the common foundations and implement the protocol when your main channel requires it.

Is Magento 2 compatible with AI agents? Yes: Magento already exposes REST and GraphQL APIs. The usual bottleneck is not the platform, it is product data quality and stock consistency.

How do I know whether ChatGPT or Gemini "see" my store? It takes targeted tests on your sector's commercial queries plus a review of crawl logs for AI user agents. We include this in the audit.

Does a Hyvä theme help? Indirectly yes: clean markup, light HTML and correct structured data make pages far more extractable than a heavy JavaScript frontend.

How long does it take to become agent-ready? With a tidy catalogue, 60-90 days. With inconsistent product data or no PIM, most of the time goes into clean-up.

Does anything change for B2B? A lot: customer-specific price lists, minimum quantities and contractual terms must be exposed in a way an agent understands, otherwise it will show wrong prices.

Where do I start on a limited budget? Product feed and structured data. Cheapest part, and it unlocks everything else.

Want to know where your store stands? Request the free technical audit.