The Protocol War That Will Decide the Future of Retail

Everyone is currently focused on the AI model war.

OpenAI vs Google. Claude vs Gemini. Who has the smartest reasoning. Who wins the interface. Who owns the consumer.

But the more important story for retail is happening several layers underneath that conversation, inside a protocol war that most retail leadership teams are barely discussing yet.

MCP. A2A. UCP. AP2.

The names sound technical and obscure today, in the same way APIs, cloud infrastructure and app ecosystems once did to traditional retailers fifteen years ago.

But these protocols are quietly becoming the foundation for how AI agents will discover products, access inventory, compare prices, negotiate preferences, authenticate trust and eventually transact on behalf of consumers.

And that changes the strategic question for retailers completely.

From mobile-first to agent-first commerce

For the last two decades, the major digital commerce decisions looked something like this. Should we sell on Amazon. Should we invest in Shopify. Should we optimise for Google. Should we build an app. Should we join marketplaces. Should we integrate with TikTok Shop.

Those decisions defined visibility, traffic and ultimately revenue.

But the next strategic decision may look very different. Which AI agent ecosystems should we expose our catalogue to.

Because in an agent-driven world, discoverability may no longer primarily depend on brand advertising, search rankings or mobile app engagement. It may depend on whether autonomous agents can actually access, interpret and trust your systems.

That is a profound shift. We are potentially moving from mobile-first commerce to agent-first commerce. And most organisations have not fully internalised what that means operationally or strategically.

Protocol compatibility is the new shelf space

Historically, every major commerce shift has ultimately been about controlling distribution and discoverability. Physical retail fought for shelf space. Ecommerce fought for SEO rankings. Marketplaces fought for seller concentration. Mobile fought for app installs. Social commerce fought for algorithmic reach.

Now we are entering a phase where discoverability itself may become protocol-driven.

If an AI shopping agent cannot easily access your inventory, understand your product structure, validate your pricing, negotiate your policies or complete transactions through the protocols it supports, your products may never even enter the consideration set. Not because consumers rejected you. Because the agent never surfaced you.

That is the uncomfortable implication. Retailers may spend the next five years investing heavily in AI-generated content, conversational commerce and frontend experiences while completely underestimating the strategic importance of the infrastructure layer underneath it all.

The winners in this next era may not necessarily be the retailers with the best websites, the biggest media budgets, the strongest apps, or even the most recognisable brands. The winners may simply become the easiest for AI agents to interact with. Easy to discover. Easy to query. Easy to trust. Easy to transact with. Easy to integrate into autonomous buying systems.

Protocol compatibility may become the new shelf space.

The four protocols that matter

The emerging stack is starting to take shape, and the progression tells a clear story.

MCP, the Model Context Protocol, is how agents access tools and data.

A2A, the Agent to Agent protocol, is how agents communicate and coordinate with each other.

UCP, the Unified Commerce Protocol, is the commerce orchestration layer that lets agents work across catalogues and merchants.

AP2, the Agent Payments Protocol, is where agents can actually authenticate identity, execute payments and complete transactions. AP2 is a Google-led extension of A2A, launched in September 2025 and backed by Adobe, Adyen, American Express, Airwallex and others.

Without AP2, the agent commerce conversation stays theoretical. With AP2, the endgame becomes visible. Agents are not just recommending products. They are completing purchases autonomously.

That is the moment retail moves from AI-assisted commerce to agent-executed commerce.

The board-level questions

Once retailers begin exposing catalogues directly to agent ecosystems, the power dynamics of commerce start changing again. The questions boards may eventually need to answer include the following.

Which ecosystems get access to our catalogue. Which agents are allowed to transact. What data do we expose. Who controls ranking logic. How do we protect margin visibility. What happens when agents optimise purely for price and convenience. Does brand loyalty weaken if agents become the primary shopping interface.

These are not technology questions anymore. They are distribution strategy questions, the same way marketplace strategy became a board-level discussion during the Amazon era.

The new control point

There is one more implication that does not get discussed enough. The protocol layer may quietly become the next major control point in commerce.

The companies that define the standards for how agents interact with retailers may end up influencing discovery, trust, transaction flow, recommendation logic, interoperability, and ultimately economic value capture.

That is why the protocol war is the real story. Not because protocols themselves are exciting, but because infrastructure layers have historically created some of the most powerful winners in technology and commerce.

The moment agents can not only discover products, but also authenticate identity, execute payments and complete transactions through protocols like AP2, retail stops being purely a human-interface industry and starts becoming a machine-interface industry as well.

Most people only notice those shifts once the market structure has already changed around them.