The playing field is mapped: What the Agentic Commerce Ecosystem Map reveals about the state of play

There comes a moment in every tech wave when someone starts drawing a map. Not because everything is clear—but because enough players have emerged that it’s hard to keep track of them all. For cloud computing, that moment came in 2009; for headless commerce, sometime in 2020. For agentic commerce, that moment has arrived.

M Farah, the author of the “Agentic Commerce Frontier” newsletter, has published an ecosystem map—and updates it monthly. In March alone, over 30 new logos were added, ranging from payment providers like InFlow Pay to marketing platforms like Klaviyo and procurement specialists like Fairmarkit. The map is getting denser. Fast.

What does the map show—and what doesn’t it show?

The Ecosystem Map is organized along the shopping journey: Discovery, Consideration, Conversion, Fulfillment, Post-Purchase. Added to this is a merchant infrastructure layer for platforms, protocols, and agent orchestration. In total, there are over a dozen categories, ranging from “Search & Discovery” to “Agent Platforms & Orchestration.”

What immediately stands out: The “Conversational & Customer Service” and “Marketing & Engagement” categories are the most densely populated. Cavalry, Flip, KODIF, Tidio, Klaviyo, Braze—this is where the crowd is. That’s hardly surprising: Customer service bots and marketing automation are the low-hanging fruit. They already work today, with existing infrastructure.

Things thin out under “Protocols & API Standards” and “Agent Platforms & Orchestration.” That’s exactly where the truly exciting questions lie: Who coordinates multiple agents? What language do they use to communicate with each other?

Three layers — a different perspective

In our three-layer model for Agentic Commerce, we distinguish three levels: AI-powered search (relevant today), AI-powered checkout (undergoing transformation), and autonomous shopping agents (the future starting in 2028). The Ecosystem Map can be viewed through this lens—and doing so reveals a lot.

Layer 1 dominates the map. Most of the listed companies are improving search, recommendations, and personalization—with AI support. Doofinder, Clerk.io, Aampe: These are tools that optimize the existing purchasing process. Humans still make the final decision themselves.

Layer 2 is sparsely populated. Payment providers like InFlow Pay, Sapiom, or DEUNA appear, but the big question—who controls the checkout when an agent makes a purchase?—remains unanswered. Spreedly has positioned itself here with the argument that merchants must retain Merchant-of-Record control while AI agents orchestrate in the background.

Layer 3 exists as a category, but hardly as a reality. "Agent Platforms & Orchestration" has one entry: Dataiera. That’s it. Autonomous agents that independently negotiate prices and trigger orders? The map honestly shows that we’re still a long way from that.

The truly exciting signals are coming from elsewhere

While the map counts logos, the interesting developments are happening in the weekly reports. In the same March issue of Agentic Commerce Frontier, Farah reports on concrete progress: Mirakl and J.P. Morgan Payments are integrating catalog orchestration with payment infrastructure. Santander and Visa have completed pilot transactions with AI agents in five Latin American countries. Azoma has launched a new standard with the "Agentic Merchant Protocol" — supported by Mars, L'Oréal, and Unilever.

The pattern: It’s not the startups on the map that are setting the pace. It’s the incumbents—Visa, J.P. Morgan, Amazon—who are making their existing infrastructure agentic-ready. Amazon is expanding “Shop Direct” via third-party feeds and building out “Buy for Me.” The strategy is clear: Agents should run on Amazon’s tracks, not on others’.

What’s missing from the map?

Three blind spots stand out:

First: Europe. The map is heavily US-centric. Where are the European commerce platforms? Where are the DACH players? This may be due to the author’s perspective, but it also reflects a reality: agentic commerce is currently being tested primarily in the US and Latin America.

Second: The protocol level. Google’s Universal Checkout Protocol, the Agentic Merchant Protocol launched by Azoma, Anthropic’s Model Context Protocol—the infrastructure through which agents communicate with commerce systems is underrepresented on the map. Yet that is where the battle for control is decided.

Third: The question of trust. “Trust & Risk” has a single entry: Riskified. But when autonomous agents make purchases on behalf of consumers—who is liable? Who authorizes? This is not just a technical question, but a regulatory one, and the answer will differ in Europe from that in the U.S.

The map is not the territory

Ecosystem maps are useful because they bring order to chaos. They are dangerous when mistaken for reality. The Agentic Commerce Ecosystem Map shows that an ecosystem is emerging. It also shows that this ecosystem is still heavily anchored in Layer 1—better search, better recommendations, better customer service.

The question that’s on my mind: Will the players on this map still be the relevant ones in two years—or will Visa, Amazon, and Google simply take over the field because they control the infrastructure?