The Agentic Commerce Stack: Why the Protocol War Is the Wrong Strategic Frame
There is a debate consuming enterprise technology circles right now about which AI commerce protocol will emerge victorious. MCP, A2A, ACP — the acronyms multiply weekly, and so do the think-pieces declaring one winner and the rest losers. This framing is wrong, and leaders who accept it are asking the wrong question.
The protocol debate matters, but not in the way it is being framed. The real strategic question is not which protocol wins — it is who controls the intelligence and ranking layer that sits above all of them. That is where market share will actually be won and lost in the next three to five years.
From Click-to-Buy to Delegate-to-Buy
For three decades, commerce operated on a behavioural model where every action required a human keystroke. A query typed, a results page scrolled, a product clicked, a checkout completed. The entire funnel assumed human initiation at every step.
That assumption is breaking. AI agents now conduct research, compare options across catalogues, apply personal preferences, and execute transactions without the human touching a browser. Conversational search is already generating higher engagement and broader product consideration sets than keyword search — particularly for lower-consideration purchases where the research overhead matters most.
This is not replacement; it is addition. Physical retail has grown at roughly 3% annually for decades while e-commerce grew at 6%. New channels augment rather than destroy what came before. Agentic commerce will follow the same pattern — but the brands that build for it early will capture a disproportionate share of the new channel before the late movers arrive.
Why the Protocol War Frame Fails
The "protocol war" narrative treats emerging AI commerce standards as competing alternatives fighting for dominance, the way VHS and Betamax or Blu-ray and HD-DVD fought. That analogy is wrong.
The emerging protocols address different problems and occupy distinct layers of the commerce infrastructure stack. Discovery protocols handle how agents find and read product catalogues. Transaction protocols standardise how agents execute checkout. Identity and consent protocols verify who an agent is acting on behalf of and what it is authorised to do. Payment and settlement protocols integrate with existing financial rails.
These are not competing solutions to the same problem. They are complementary components of a stack, like TCP/IP, HTTP, DNS, and SSL on the modern internet. Nobody declared HTTP the winner of the "internet protocol war" — they all coexist because they solve different problems at different layers. The same is true of the agentic commerce stack. Multiple standards will emerge, coexist, and interconnect. Enterprises that wait for a single winner before engaging will wait too long.
The Five-Layer Commerce Stack
A more useful strategic frame than "protocol war" is architectural. The agentic commerce stack has five distinct layers, each with its own competitive dynamics:
Layer 1 — Discovery and Intent. How agents find and read product information. The key requirement here is structured, machine-readable catalogue data. Agents cannot scrape unstructured pages reliably. Brands with rich, accurate, API-accessible product data will be discoverable. Those without it will be invisible to agent-mediated commerce, regardless of which discovery protocol dominates.
Layer 2 — Action and Checkout. Standardised protocols that allow agents to execute transactions while preserving merchant control over pricing, promotions, and terms. This layer is where the major platforms are competing most visibly — but it is a lower-stakes battle than it appears, because the transaction mechanics will largely commoditise.
Layer 3 — Identity, Consent, and Trust. Cryptographic verification of agent identity and authorisation. Who is this agent acting for? What has the user actually authorised it to do? This layer solves the fraud and liability questions that will otherwise stall enterprise adoption. It is unsexy infrastructure, but it is the foundation everything else depends on.
Layer 4 — Payments and Settlement. Integration with existing financial infrastructure — card networks, wallets, B2B payment rails. The innovation here is largely in the orchestration, not the underlying payment mechanisms. Existing payment processors will adapt; this is not where differentiation lives.
Layer 5 — Intelligence and Ranking. This is the layer that actually matters strategically. It determines which products agents recommend, in what order, weighted by what signals. Small changes in ranking algorithms drive outsized shifts in market share. And unlike the lower four layers, which will eventually standardise and commoditise, intelligence is permanently differentiated. It compounds over time. Enterprises that build proprietary intelligence infrastructure early — rich behavioural data, personalisation models, demand forecasting — create durable competitive moats.
The Real Battleground
Layer 5 is where the battle for agentic commerce will actually be decided. The protocol debates in Layers 1 through 4 are important infrastructure questions, but they resolve into standards over time. Ranking does not resolve into a standard — it remains a proprietary capability that depends on the quality and uniqueness of your data.
This has a direct implication for enterprise strategy: the priority in the next 12 to 24 months should not be picking a protocol winner. It should be building the data infrastructure — structured product data, behavioural signals, personalisation models, analytics pipelines — that will power superior intelligence at Layer 5, regardless of which lower-layer protocols emerge.
"Small changes in ranking algorithms can drive outsized shifts in market share." This will be as true in agent-mediated commerce as it is in search today. The enterprises that own their ranking intelligence will own their visibility. Those that outsource it to platform defaults will compete on the platform's terms.
What Enterprise Leaders Should Do in the Next 12–24 Months
First, invest in structured product data immediately. This is the prerequisite for discoverability at Layer 1, regardless of protocol. Rich, accurate, machine-readable catalogue data is the single highest-priority capability investment for any commerce enterprise in 2026.
Second, build for multi-protocol interoperability rather than betting on a single standard. Composable commerce architecture — decoupled, API-first, modular — allows enterprises to participate in multiple ecosystems simultaneously without full-stack rewrites. This is the right bet given genuine protocol uncertainty.
Third, maintain ownership of intelligence and analytics. Do not outsource ranking to platform defaults. Invest in the proprietary data infrastructure — behavioural signals, personalisation models, demand forecasting — that will generate durable advantage at Layer 5.
Fourth, establish governance frameworks that demand transparency in ranking practices from platform partners. If your products are being surfaced to AI agents through a third-party platform, you need to understand and influence the ranking logic. Opaque ranking is a strategic risk.
Fifth, do not wait. Delaying engagement with agentic commerce infrastructure means losing visibility in AI-mediated discovery channels as they scale. The window for early-mover advantage is open now. It will not remain open indefinitely.
B2B: A Different Kind of Disruption
The consumer commerce frame dominates the agentic commerce conversation, but B2B deserves separate attention. B2B purchasing is structurally different: negotiated pricing, complex contracts, multi-level approval hierarchies, and relationships built over years. Autonomous agent checkout — the model that works for consumer low-consideration purchases — does not map cleanly onto B2B procurement.
In B2B, agents function as decision accelerators rather than autonomous transactors. They assist with specification matching, vendor comparison, compliance verification, and preliminary pricing analysis — dramatically compressing the research phase of complex purchasing cycles. But humans remain in the approval and authorisation loops, by design.
This creates a different set of infrastructure requirements. B2B catalogues need to expose negotiated pricing, contract terms, and configuration complexity in machine-readable formats. Identity and consent infrastructure (Layer 3) matters even more in B2B, where authorisation hierarchies are complex. And the intelligence layer needs to understand procurement context — budget cycles, preferred vendor relationships, compliance requirements — not just product attributes.
Enterprise software vendors and B2B distributors that build for agent-assisted procurement now will compress sales cycles and reduce friction in a way that drives real revenue impact.
The Strategic Imperative
McKinsey estimates AI-driven commerce could generate between one and five trillion dollars in annual economic value by 2030. Even at the conservative end of that range, the magnitude justifies immediate strategic attention.
But the framing of that opportunity matters. Agentic commerce is not a protocol battle to be won or lost in a standards committee. It is a new distribution channel with its own infrastructure stack — and like every distribution channel before it, the brands that build for it first, and build the right capabilities at the right layers, will capture a disproportionate share of what it generates.
The question worth asking is not which protocol wins. The question is: which enterprises will control their intelligence infrastructure when agents become the primary discovery channel — and which will be optimising for a ranking algorithm they do not understand and cannot influence.