AI traffic to U.S. retailers surged 393% in Q1 2026 alone. That is more than a traffic spike—it is a signal that agentic commerce is beginning to behave like a real channel, even as many enterprises still lack the content and transaction infrastructure required to participate in it effectively.
The Numbers Are Not a Forecast – They Are Real!
Six months of data from Adobe, Salesforce, Shopify, Adyen, and McKinsey point in the same direction: agentic commerce is no longer a fringe concept or a future-state thought experiment. It is starting to emerge as an active commerce channel—one that retail executives, CIOs, CDOs, CTOs, and digital commerce leaders should now evaluate with the same seriousness they once applied to mobile, marketplaces, and digital media shifts.
The more important shift is not simply traffic growth; it is conversion quality. In March 2025, AI-referred traffic converted 38% worse than paid search and email. By March 2026, AI-referred traffic converted 42% better than non-AI traffic, with stronger engagement and higher revenue per visit. The implication is strategic: once AI-driven shoppers can access a retailer’s products and transaction flows, they are increasingly valuable customers rather than experimental traffic.
The immediate challenge is that many enterprises are still not machine-readable or machine-transactable at the level these agents require.
Most Businesses Are Invisible to the Agents Doing the Buying
That readiness gap is already visible in the data. Adobe found that 34% of product pages cannot be read by AI agents at all, while category pages and homepages also show meaningful visibility gaps. For retail and technology leaders, this is not simply a discoverability issue; it is a channel access problem.
If an AI agent cannot read a product page, that product is effectively absent from the agentic commerce channel. This shifts the optimization agenda beyond traditional SEO alone and toward a broader model of machine-readable content, structured data, and answer-engine visibility.
How to get AI-ready
- Add structured product data – AI agents need access to your prices, stock levels and product detail labels
- Crawler access – Crawler access is table stakes, but does not guarantee a competitive advantage
- Optimize content for answer engines – Rewrite your product pages to answer questions – not just describe products
- Document your API’s with OpenAPI – Create a plain-language description of what your store can do
- Deploy a static MCP catalog – Publish a fixed, approved list of what AI agents are allowed to access in your store
- Route AI traffic through a secure gateway – Add a security checkpoint between AI agents and your store
Three Protocols Now Govern How Agents Transact
That infrastructure readiness is beginning to consolidate around a small set of open standards. For executives evaluating where to invest, these protocols matter because they define how agents discover products, access transaction capabilities, and complete purchases across platforms.
- MCP (Model Context Protocol) enables AI assistants to call structured tools against a retailer’s catalog, cart, and checkout systems. Commercially, it matters because it turns product and transaction data into something agents can act on directly rather than merely reference.
- UCP (Universal Commerce Protocol) provides a framework for agents to query catalogs and execute purchases across participating commerce ecosystems. For retailers, it expands reach into the environments where AI-mediated product discovery is increasingly happening.
- ACP (Agentic Commerce Protocol) governs in-chat product discovery and checkout flows. Its importance is practical: it supports the shift from AI-assisted browsing to AI-enabled transactions completed inside conversational interfaces.
Retailers without structured API infrastructure are effectively excluded from this layer of commerce. Supporting the protocols is what allows agents to find, evaluate, and transact with a business in a reliable way.
However, protocol support alone does not solve the underlying data problem. An MCP server only exposes what sits behind it. If catalog data is inconsistent, pricing logic is fragmented, or inventory accuracy is unreliable, those weaknesses are passed directly into agent-mediated decisions. That is why agentic readiness is not a single implementation project; it is a sequenced capability build that begins with content and data quality, then extends into governed transaction infrastructure.
Cyclotron’s Four-Phase Framework for Getting Agentic Commerce Ready
For retail executives and technology leaders, the practical question is not whether agentic commerce matters. It is how to prepare for it in a way that is sequenced, governed, and commercially useful. A four-phase framework helps clarify where to start, what each investment unlocks, and what risks emerge when steps are skipped.
Phase 1 — Readiness Assessment establishes the baseline. This phase identifies AI visibility gaps across the catalog, reviews crawler and policy settings, evaluates structured data coverage, and benchmarks performance against category peers. The business outcome is directional clarity: leaders can prioritize investment based on measurable exposure and risk. Skipping this phase often leads to fragmented spending and unclear sequencing.
Phase 2 — AI Visibility and Content Readiness closes the machine-readability gap. This work includes structured schema implementation, content rewrites for answer-engine retrieval, FAQ and attribute standardization, and improvements to how product information is surfaced across AI-facing environments. The outcome is greater discoverability and recommendability by the agents driving upstream product research. Without this phase, even strong brands risk becoming difficult for AI systems to interpret or cite.
Phase 3 — API-First Commerce and Transaction Infrastructure builds the systems agents can transact through. This includes commerce APIs designed for agent consumption, governed API management, MCP server deployment, headless checkout capabilities, and transaction controls such as spend limits, policy restrictions, and human-in-the-loop thresholds. The outcome is reliable machine transactability. If this phase is underbuilt, visibility may improve without creating a dependable path to conversion.
Phase 4 — Autonomous Commerce Operations extends the model to agent-to-agent transactions, dynamic pricing, AI-driven personalization, agent identity, fraud controls tuned to agent traffic, and fulfillment or supply-chain APIs. The outcome is a governed operating model for increasingly autonomous buying journeys. Entering this phase without identity, trust, and policy controls in place creates operational and reputational risk rather than strategic advantage.
The Phase Sequence Matters
The sequence matters because each phase depends on the integrity of the one before it. An MCP server built on fragmented catalog data only scales fragmentation. Autonomous commerce without identity and transaction controls increases exposure rather than capability. The organizations moving fastest are not ignoring sequence; they are accelerating it under a unified program architecture.
The near-term advantage will go to retailers that make their catalogs machine-readable, their transaction systems agent-accessible, and their governance models fit for autonomous commerce. For retail executives, CIOs, CDOs, CTOs, and digital commerce leaders, the strategic imperative is straightforward: treat agentic commerce as a channel that now requires readiness across content, data, APIs, and control frameworks.
About the Author
Andrew Fletcher is Vice President of Azure Innovation at Cyclotron, where he helps enterprise organizations translate emerging technology shifts into practical cloud, data, and AI strategies. He works closely with business and technology leaders to modernize digital platforms, strengthen data foundations, and prepare for new operating models shaped by AI. His focus spans innovation strategy, enterprise architecture, and the application of Azure technologies to complex transformation initiatives.