ACO Is a Function, Not a Project

The conversation about agentic commerce has shifted in the last year. Twelve months ago, senior commerce leaders were asking what AI shopping agents are and how they work. Today, after seeing how products get recommended (or excluded) by AI shopping agents across a growing number of channels, the question has moved on.

The question now is operational: who owns this inside the company, how existing teams need to coordinate, what good looks like, and how to fund it. Those are operating model questions. And they are the questions that will separate the brands and retailers that compound advantage in agentic commerce from the ones that fall behind.

For readers landing here cold: Agentic Commerce Optimization (ACO) is the discipline of governing whether your products are eligible for inclusion across AI shopping experiences, at the SKU level, across every agentic shopping engine, continuously. Our What is ACO primer covers the working definition; this series picks up downstream of it and focuses on how organizations actually run the discipline.

Why the operating model is the question now

The market is moving faster than most organizations are. McKinsey research finds that only 1% of companies believe they have reached AI maturity. BCG's recent work puts hard numbers on the gap: 5% of companies are "future-built" and seeing 5x revenue increases compared with their peers, while 60% of companies report little material AI value at all.

The window for organizational advantage is wide open precisely because most organizations have not yet closed the operating model gap. Most have started experimenting with ACO in at least one channel. Most have not yet built the function that makes the experimentation durable, or that connects what they learn in one channel to how they optimize across the rest.

ACO is a function, not a project

The strongest organizational analog for ACO is retail media. Retail media started as a marketing experiment inside a handful of large retailers. It ran as a project for a few cycles. Then it matured into a dedicated function with its own leadership, P&L, and measurement system.

Bain estimates there are now around 150 retail media networks globally, growing at roughly 12% per year with margins above 50%. The brands and retailers that built it as a function early compounded advantage. The ones that kept treating it as a campaign-by-campaign initiative spent years catching up. ACO is on the same trajectory.

That distinction matters because the failure mode is predictable. Treating ACO as a project produces a flurry of initial activity, a few wins in one channel, and then a slow drift back to status quo when the project ends. Treating ACO as a function produces something different: a continuous loop where every channel your products touch sends signal back, how agents evaluated them, what attributes mattered, where you were eligible and where you were not, and that context feeds the next round of optimization.

Over time the system gets smarter, coverage expands, and conversions improve across every channel simultaneously. That is a compounding advantage, and it only works when the operating model supports it.

The operating model is the value gap

McKinsey's most recent State of AI research found that workflow redesign has the largest impact on EBIT among all the practices that drive value from generative AI. They also found that only 21% of organizations have fundamentally redesigned at least some workflows.

The implication is direct. ACO is one of the places where that gap can be closed in a contained, measurable way, because the operating model questions ACO raises (who owns it, how teams coordinate, what gets standardized, what gets exception-reviewed) are the same questions that determine whether any agentic AI initiative produces value.

McKinsey's recent work on the agentic organization puts a finer point on this. Organizations that succeed with agentic AI align across five dimensions: business model, operating model, governance, workforce and culture, and technology and data. ACO requires that same alignment, applied specifically to product discovery and the growing number of AI-driven channels where buying decisions now happen.

The five layers of an ACO operating model

We use five layers to describe how a mature ACO function operates. They are loosely parallel to McKinsey's framing for the agentic organization, but specific to commerce. Most organizations are strong in one or two of these layers today. The companies moving fastest align all five.

The strength of an ACO function is not measured by which layers exist in isolation. It is measured by how aligned they are. Strategy without ownership produces decks no one acts on. Ownership without workflow produces a single person who cannot scale. Workflow without governance produces inconsistency. Governance without systems produces bottlenecks. Each layer reinforces the next, and the function compounds only when they reinforce each other in both directions.

Where this goes next

The five-layer model is a frame. It does not tell you where your organization stands today. That is the work of Part 2.

The ACO Maturity Model uses these same five layers to identify three stages, Explorer, Builder, and Leader, that map to how most commerce organizations actually operate. Part 2 introduces a short interactive assessment that places your organization in the right stage based on how the layers are functioning today, and lays out what the move to the next stage looks like.