How to Build Your ACO Team

Part 1 argued that ACO is a function, not a project. Part 2 introduced the maturity model and a short assessment to diagnose where your organization sits today. This is the operational close: what a mature ACO function actually looks like when it runs.

The shorthand: name the owner, coordinate the six disciplines that already exist somewhere in your organization, govern with guardrails rather than gatekeeping, and put a system layer behind it so the work scales. Each of those four moves is what separates the organizations compounding advantage in agentic commerce from the ones still treating it as a side project.

The single-threaded owner is non-negotiable

Every successful ACO implementation we have seen has one person who owns it end-to-end. Senior enough to make decisions across functional boundaries. Close enough to execution to understand a fast-moving landscape. Every struggling implementation lacks this.

The most useful description of why fragmented ownership fails comes from Harvard Business Review's analysis of "agent sprawl." When decentralized teams spin up AI capabilities without a unifying strategy, the result is duplicative work, security risk, and a collection of disconnected point solutions instead of a cohesive system. ACO is particularly vulnerable to this pattern because the work touches catalog, brand, SEO, legal, merchandising, and engineering. Without an owner, every team optimizes locally.

The single-threaded owner is typically drawn from VP or Director of E-Commerce, Head of Digital, or Senior Director of SEO. The exact title matters less than three things. They have enough seniority to make decisions across functional boundaries. They are close enough to execution to understand how product data and AI shopping agents actually behave. And they have the authority to set policy on behalf of the organization, not just propose it.

This is not a new headcount in most cases. It is a named accountability assigned to someone who already has adjacent responsibility.

Six disciplines, not a new department

A common reaction to "ACO needs a function" is "we cannot afford to build a new department." The good news is that you almost certainly do not need to. The disciplines required for ACO already exist in most commerce organizations. What changes is how they coordinate.

Six disciplines underpin a mature ACO function. Think of these as practice areas, similar to retail media when it was emerging, not an org chart you copy. One person can cover multiple disciplines at smaller organizations. At enterprise scale, each becomes a dedicated role or small team.

These disciplines are usually distributed across existing teams. At Explorer stage, one person covers three or four. At Builder, a small team of two to four handles the core disciplines with content review shared. At Leader, each becomes a dedicated owner. The disciplines do not change as the organization scales. The staffing does. This mirrors how digital marketing matured: one person ran SEO, paid search, and analytics, and as the function grew, each became a dedicated role.

Governance through guardrails, not gatekeeping

Governance is where most organizations either accelerate or stall. The instinct is to slow ACO down with traditional approval chains. The result is a function that cannot operate at catalog scale and an organization that gradually disengages from the work.

PwC's 2026 AI Performance Study found that leading AI performers are roughly twice as likely to redesign workflows around AI and 2.8 times more likely to increase decisions made without human intervention. They are also going further on governance than their peers. The two move together.

The shift is from gatekeeping to guardrails. In a gatekeeping model, every product change requires individual approval from legal, brand, product marketing, or merchandising. That works at small scale. It does not work when the catalog has thousands of SKUs and AI shopping agents change their evaluation logic regularly.

In a guardrail model, the organization makes two distinct types of decisions:

The multi-stakeholder reality at enterprise scale makes this distinction essential. Large brands and retailers do not have one person approving product data changes. They have legal, product marketing, brand, and merchant teams, each with different approval criteria. Guardrail-based governance is what makes that complexity manageable: each stakeholder defines standards once, the system enforces them consistently, and human review concentrates on exceptions and edge cases rather than every SKU.

This approach respects the legitimate concerns that brand, legal, and compliance teams bring to product data. NRF research consistently finds that retailers are adopting AI quickly while remaining cautious about risk, accuracy, cost, and regulatory complexity. That caution is reasonable. Guardrail-based governance honors it without creating a bottleneck that prevents the function from scaling.

The system layer makes it work at catalog scale

Operating model alignment is the strategic answer. Continuous execution is the practical one. Without a system layer, every part of the ACO function falls back into manual work that does not scale.

A recent survey found that more than 80% of retailers see agentic AI as a path to efficiency, but data, integration, and scaling are the most cited challenges. That gap between intent and execution is what a system layer closes. The system is responsible for evaluating product data against agentic commerce requirements at the SKU level, identifying gaps and prioritizing enrichment opportunities, managing enrichment queues, distributing optimized product data across every channel where AI-driven discovery happens, and monitoring coverage and performance continuously.

Critically, the system also captures context back from those channels: how agents evaluated your products, which attributes drove eligibility, where competitors appeared alongside you, what changed since the last cycle. That context feeds directly into the next round of optimization. This is what turns ACO from a push operation into a learning loop. Context flows in, optimized catalog data flows out, and the system gets smarter with every cycle.

This is not a black box. The disciplines stay accountable for what gets approved, what gets monitored, and what gets prioritized. The system makes that work possible across the full catalog instead of a curated subset. It also creates the audit trail and consistency that governance requires.

The window is now

A recent Harvard Business Review analysis of agentic AI transformations found that 74% of executives whose organizations introduced agentic AI saw returns within the first year when those organizations approached the work as a system rather than a collection of point projects. The gap between leaders and laggards in agentic commerce will widen quickly.

Most organizations are still Explorers. The ones that move to Builder this year will have the operating model in place to compete in the 2026 holiday cycle and the 2027 transactional surge. The ones that wait will be optimizing for channels that have already moved past their products.

The compounding effect matters. Every month an organization runs the loop, context in, optimized catalog out, performance measured, lessons fed back, the system gets smarter and the coverage gap with competitors widens. Every month they do not run it, they are not standing still. They are falling behind organizations whose flywheel is already turning.

ACO is not a new department to build. It is a function to name, an owner to assign, six disciplines to coordinate, and a system that lets the work happen across every channel where AI-driven commerce is taking hold. The companies that treat it that way will compound advantage. The companies that treat it as a project will keep finding themselves three steps behind.

The move from Explorer to Builder is achievable in 90 days when an organization commits to it. The first step is always the same: name the owner.