Last updated on February 12th, 2026 at 03:08 am

update 2/12: (screencap from Twitter/X)

Google’s recent preview of shopping ads inside AI Mode provides a visible example of this restructuring already underway. Framed publicly as a new advertising format, the more meaningful shift is architectural. Comparative merchant offers now appear directly within conversational responses, with price, fulfillment, and policy attributes rendered as structured objects inside dialogue. Discovery is no longer confined to ranked links; it is mediated through an interpretive layer that assembles and presents offers in response to articulated intent.

What remains largely invisible to users is the coordination layer beneath that surface. Merchant feeds, structured product graphs, ranking systems, and emerging commerce protocols operate below the conversational interface. Negotiation, constraint resolution, and offer matching occur at layers abstracted away from view. The interface appears fluid and simplified precisely because orchestration has migrated downward into protocol and data layers. This is how agent-mediated commerce becomes operational: not through dramatic interface reinvention, but through the quiet relocation of transactional logic beneath conversation itself.

original post:

Read the business version

Shopify, Google, and over 20 partners announced a new agentic-AI open ecommerce protocol today called UCP (Unified Commerce Protocol). The protocol reflects a shift toward highly segmented, scaffolded agent architectures designed to aggressively manage session limits and hallucination risk. Rather than embedding AI into existing storefronts, UCP enables commerce to occur wherever intent is expressed or outreach is possible: chat interfaces, AI assistants, messaging systems, or other conversational surfaces. Transactions no longer depend on HTTP or web navigation, and commerce logic is decoupled from websites entirely. This is likely to make some elements of ecommerce more visible, and others less transparent.

This architecture introduces a structural change across the ecommerce value chain, borrowing heavily from B2B procurement systems rather than consumer retail models. AI is no longer positioned as just a recommendation layer, but as a coordinator that translates intent into requirements and resolves fit throughout the purchase process. There are technical, buyer, and business owner benefits. The result is a new, highly distributable execution pathway.

From Browsing to Requirement Fit

Let’s look at a practical example. A user tells their AI assistant they want to fly from Toronto to London within the next two weeks, prefers evening flights, needs an aisle seat, and wants to stay under $800 CAD. The agent translates this into structured requirements and queries airline or intermediary systems. Those systems respond with conditional offers based on inventory and pricing rules. The agent may explore alternatives within allowed bounds (different dates or routes), but it cannot invent discounts or override requirements. Once feasible options are identified, the user approves one, and payment and booking execute through existing airline or intermediary systems.

Airlines retain authority over inventory and pricing. Intermediaries normalize data. The AI coordinates and explains. The user approves execution. What disappears is manual searching, comparison across tabs, and multi-step checkout flows. Websites remain relevant for brand trust and policy details, but they no increasingly, with adoption, do not serve as the transaction engine.

This is not speculative. It is how many agent-mediated commerce protocols are being structured now, importing enterprise procurement logic into consumer transactions.

From ACP to UCP

The Agentic Commerce Protocol (ACP), launched in September 2025, is another open-source agentic protocol led by OpenAI, Stripe, and a group of payments, commerce, and platform partners including PayPal to define how AI agents securely execute transactions on a user’s behalf. ACP focuses specifically on the transaction layer (authorization, payments, confirmations, and settlement) rather than discovery or orchestration. This distinguishes it from UCP, which addresses how agents identify options and coordinate purchasing decisions across vendors. In practice, ACP does not change how merchants run their businesses or how storefronts operate; it standardizes how an already-chosen purchase is completed when an agent, rather than a human, initiates the transaction.

The Core Architectural Shift

UCP is a broader, more ambitious enablement platform than ACP. The operational shift is from catalog presentation to transaction layer to cross platform, distributed coordination and requirement negotiation. Traditional e-commerce displays options and relies on users to determine fit. Agent-mediated systems resolve feasibility first, presenting only valid options.

The process is deliberately segmented into discovery, negotiation, execution, and verification, each with clear authority boundaries. AI systems translate intent and orchestrate communication, but do not set prices, hold inventory, or commit funds. Those functions remain with merchant, payment, and fulfillment systems. This separation is what allows agent-mediated commerce to scale safely.

Architecturally, these protocols resemble application-layer messaging standards more than platforms. Commerce logic is expressed through machine-readable messages describing capabilities and requirements, not through page navigation or checkout flows. Decision-making moves into conversational interfaces, while authority remains external.

Why This Model Exists

This open-source structure is derived directly from B2B procurement and it means everyone can now build on the same protocol in the same way, while making the process distributable to anywhere a buyer is looking for information. Enterprise purchasing has always relied on explicit requirements, negotiated terms, and rule-based authority. Pricing is conditional, not static. Agent-mediated commerce adapts this logic to consumer contexts to make complex transactions machine-navigable.

Because these patterns are well established, the primary challenges are adoption and standardization, not technical feasibility.

Control and Accountability

Economic commitments are made only by authoritative systems, not generative output.

Merchants control inventory, pricing rules, and fulfillment.

Payment providers control funds and compliance.

Intermediaries normalize data and often act as the counterparty.

AI systems orchestrate, translate, and explain, but cannot execute independently.

Users specify intent and approve final execution.

Fixed-Price Vendors Still Drive

This model also accommodates small vendors without dynamic pricing. A handwoven wool blanket maker with fixed prices, limited inventory, and simple shipping rules does not need sophisticated systems. Their requirements are driven by circumstance: fixed price, next shipping date, no alternatives.

The agent identifies the vendor because its product matches the buyer’s stated needs, not because the vendor paid for placement. Visibility comes from attributes and availability, identified through structured data or intermediaries, rather than marketing scale.  In most cases, this does not require any new systems or changes. Clear public information, a standard website, basic product descriptions, and accurate availability is enough for discovery, either directly or through existing platforms and intermediaries.

What Changes, and What Doesn’t

Agent-mediated commerce shifts where effort sits in the commercial stack rather than removing layers altogether. Tools designed to handle consumer-side discovery and product comparison become less central, while inventory management, fulfillment, payments, and accounting continue to anchor the transaction.

Websites are still critical, but primarily as informational and trust-building assets. Execution moves into agent-coordinated flows embedded in chat, voice, or messaging environments.

Where Adoption Happens First

Early adoption will concentrate in sectors with conditional pricing, dynamic availability, and high discovery cost: travel, logistics, insurance, ticketing, professional services, and B2B procurement. These sectors already operate with rule-based systems and intermediaries, making them structurally compatible.

By transaction complexity, these categories represent a majority of e-commerce activity. Adoption here is likely within 3–5 years.

Commoditized goods with fixed pricing see less change. Agents may assist with reordering, but pricing and architecture remain largely unchanged.

Discovery Shift and Organic Authority

Discovery will likely remain pinned to search infrastructure, particularly Google as a partner in this protocol, but the nature of discovery changes. Agents query based on structured data and relevance signals rather than visual ad placement. This shifts value toward organic authority (content quality, structured markup, inbound links) and text-based ads, easily transferable to AI conversational commerce, away from click-based advertising models. Whether intermediaries will monetize agent queries through new fee structures or maintain ad-based revenue through restructured placements remains unresolved, but the shift away from discovery via human-facing ad formats toward suggestion is coming. where social commerce failed to pan out as many had expected and predicted agentic e-commerce is structural, and its impact will likely be significantly different.

Who Gains and Who Loses

Sectors with complex feasibility gain efficiency and conversion. Procurement-heavy and service businesses benefit from clarity over persuasion. Small producers gain discoverability based on fit rather than spend. Intermediaries with clean APIs and execution guarantees become more central.

Businesses that depend on or provide services that focus on attention capture, funnel optimization, or advertising-driven visibility lose some leverage. Dark-pattern UX and psychological pricing tactics are ineffective against agents.

Large marketplaces face mixed outcomes. Those that provide execution, logistics, or trust infrastructure remain relevant; pure traffic aggregators risk disintermediation.

The Emerging Power Shift

It remains to be seen how this will manifest. Agent-mediated commerce may reverse traditional data asymmetry; it may also worsen. Merchants expose requirements data to agents, while user intent remains private until execution. This ideally limits behavioral price discrimination and increases user leverage, though it will likely reduce the visibility and number of options available. The impact of regulatory and competitive pressures remain to be seen.

Clarification, Not Disruption

This announcement represents a potentially enormous shift in how e-commerce occurs. There will likely be competitive models. UCP is open, but its execution effects will likely be proprietary, particularly in product and offering discovery given Google’s involvement.