Agentic Commerce — How AI Agents Buy on Behalf of Customers and What It Means for Your Online Store

What is Agentic Commerce

Agentic commerce is an online retail model where an AI agent handles the entire purchase process — from search and comparison to payment — on behalf of the user, without them visiting a site or clicking manually.

The difference from previous AI tools is fundamental:

The user sets parameters. The agent executes.

Why 2026 is the tipping point

Three forces converged at once to turn agentic commerce from concept into reality:

• ✓
  Mass consumer adoption. According to an IBM survey from January 2026, 45% of consumers already use AI for at least one step in the purchase process. The habit is formed — people ask AI instead of searching on Google.
• ✓
  Ready infrastructure. ChatGPT, Google AI Mode, Amazon Rufus, Perplexity — all have launched commerce features with real payment. The ACP and UCP protocols let merchants integrate once and be accessible from multiple AI platforms.
• ✓
  Open standards. Without standards, agents cannot compare products from different merchants. ACP (OpenAI and Stripe) and UCP (Google) solve exactly this problem — they create a common language between AI agents and commerce systems.

How the process works technically

The user types in ChatGPT: "Find me waterproof hiking boots under 200 euros, size 42, with delivery by Thursday."

Here's what happens next:

1. Intent parsing. The AI model understands the structured criteria — product type, budget, size, delivery deadline.
2. Request to product APIs. The agent doesn't browse sites visually. It sends structured requests to merchant APIs supporting ACP or UCP. From them it receives structured data — price, availability, parameters, delivery timeframe.
3. Comparison and ranking. The agent compares results against the given criteria. The merchant with more accurate data, better availability and shorter delivery time wins.
4. Transaction completion. With ACP, payment goes through Stripe with tokenised data — the agent never sees the user's card. With UCP, Google processes the transaction.
5. Confirmation. The user receives confirmation directly in the chat. The merchant receives the order.

The new reality: your backend is the new storefront

In a traditional online store, the customer sees your design, banners and promotions. You persuade them with visuals and content.

In agentic commerce, the AI agent sees none of that. It only sees the structured data you've made accessible — and decides whether to include you in the results or not.

If the agent cannot verify whether a product is in stock today and whether it can be delivered by Thursday — it simply chooses the competitor who can.

What your online store needs to prepare

1. Structured product data

This is the foundation. AI agents don't rely on text descriptions — they rely on machine-readable structured data.

Schema.org markup is mandatory. Every product must have correctly marked: name, price, availability, category, parameters, image, brand, identifiers (GTIN, MPN).

For technical products — power supplies, electronics, components — parameters are especially critical. The agent must be able to verify "24V, 100W, DIN-Rail form factor" programmatically, not from a text description.

2. Real-time availability

"On request only" and "select quantity" are not enough for an agent receiving a machine request. Availability must be updated in real time and accessible via API or structured data.

Agents carrying a deadline task automatically exclude from comparison any merchant where they cannot confirm availability and delivery time.

3. API layer for product data

For ACP and UCP compatibility, the merchant must maintain API endpoints allowing agents to make structured requests. This is not mandatory from day one, but it is necessary for full participation in agentic commerce.

The good news: Shopify, WooCommerce and OpenCart already have or are developing ACP/UCP integrations. For custom platforms — this requires additional development.

4. Fast and reliable checkout

Agentic commerce is built around automated transactions. A complex multi-step checkout requiring manual intervention is incompatible with the agent model.

A simplified checkout — with minimal steps, Stripe or equivalent support, clear API endpoints for orders — is mandatory for participation.

5. Reliable delivery information

In agentic commerce, delivery time is a competitive factor before purchase, not after. Agents compare "can it be delivered by Thursday" before choosing a merchant.

Integration with courier APIs for actual delivery times — not just "3–5 business days" — gives a competitive advantage in agent ranking.

B2B and agentic commerce — especially high potential

In B2B commerce, agentic commerce has even greater potential than in consumer markets. Forrester predicts that 20% of B2B sellers will face agent-led quote negotiations by the end of 2026.

The reasons are clear: B2B buyers place recurring orders to exact specifications. An AI agent that knows the parameters of the required component and approved suppliers can automate the entire procurement process — without human intervention.

For B2B online stores with technical products — like MEANWELL.bg or similar — agentic commerce is an especially relevant topic. Buyers already know what they need. The agent just needs to find the right merchant with the right data.

The ACP and UCP protocols — in brief

Most merchants will need to support both protocols — ACP for the ChatGPT audience, UCP for the Google audience.

What agentic commerce is NOT

It's important to distinguish it from similar concepts:

• ✗
  It's not a chatbot with a "buy" button. If the conversational interface redirects the buyer to your checkout — that's conversational commerce. The agent did not complete the transaction.
• ✗
  It's not a recommendation system. Amazon "customers who bought X also bought Y" is a recommendation. Agentic commerce is the agent completing the purchase.
• ✗
  It's not browser automation. Early demonstrations of "AI shopping" relied on AI visually clicking on websites. That is fragile, slow and blocked by fraud systems. Mature protocols work with APIs.

The risks of inaction

Traditional SEO optimises for people searching on Google and choosing from 10 blue links. In agentic commerce, the agent doesn't return 10 links — it chooses one.

If your product data isn't structured correctly, if availability isn't real-time, if you don't have ACP/UCP support — the agent simply skips you in favour of a competitor who does.

It's important to note: agentic commerce won't replace the traditional online store overnight. The transition is gradual. But the trajectory is clear — and the first merchants who prepare the right infrastructure will have a structural advantage that's hard for slower competitors to catch.

The new metric: AI Citation Rate

In traditional SEO, the key metric is Click-Through Rate (CTR) — how many people click your result in Google.

In agentic commerce, the key metric is AI Citation Rate — whether the AI agent includes your product in results for a query. If the agent doesn't return you, your brand simply doesn't participate in the transaction.

This metric is still hard to measure — the protocols are new and attribution is still evolving. But the direction is clear: traditional SEO metrics are gradually complemented by a new category — visibility to AI agents.

The related concept is GEO (Generative Engine Optimization) — optimisation for generative AI models, analogous to SEO for search engines. In agentic commerce, GEO concretises into optimisation of product data for agent ranking.

Parallel with the mobile revolution

When mobile devices went mainstream around 2010–2012, most websites weren't ready. The small screen, slow connection and different user behaviour required fundamental changes. Sites that adapted early gained traffic. Others suffered significant ranking losses.

Agentic commerce is the next similar shift. The scale is different — not a change of device, but a change of the agent performing the interaction. But the logic is identical: platforms prepared for the new agent will gain access to a new channel. Unprepared ones will remain invisible to it.

The timeline for the mobile revolution was 3–5 years from "early adopters" to "mass necessity". For agentic commerce — given the pace of AI development — the horizon is likely shorter.

Platform-specific readiness

Action plan

Short term (now)

• ✓
  Audit schema.org markup — are all products correctly marked?
• ✓
  Update availability in real time
• ✓
  Simplify the checkout process
• ✓
  Add accurate delivery timeframe information

Medium term (next 3–6 months)

• ✓
  Check for Shopify/WooCommerce ACP integrations (where ready-made plugins exist)
• ✓
  For custom platforms — plan an API layer for product data
• ✓
  Structure product parameters for machine readability

Long term

• ✓
  Full ACP and UCP compatibility
• ✓
  Integration with courier APIs for actual delivery times
• ✓
  Monitor agent-generated orders as a separate channel

Common questions about agentic commerce

Conclusion

Agentic commerce is not a distant trend — it is an active reality with working protocols, hundreds of millions of users and trillions of dollars in projected volume.

Most online stores in Bulgaria are still invisible to AI agents. Not because they lack good products, but because their data isn't structured for machine reading.

Preparation doesn't require a full restructuring — it requires the right technical foundation. And the earlier it's built, the greater the advantage.

Singularity Edge Studio — web development and technical optimisation for online stores in Plovdiv. Get in touch →