What an AI Center of Excellence Actually Needs to Work

A lot of organizations are building AI Centers of Excellence right now. Some have a team. Some have a mandate. Some have a Slack channel and good intentions.

What most of them are missing is the infrastructure behind the CoE itself.

An AI CoE is not a department or a job title. It is the organizational system that governs how AI gets adopted, scaled, and sustained across the business. And building one that actually works requires five interconnected infrastructure systems, each responsible for a different dimension of transformation.

Here is what those five systems are, and why each one matters.

1. Workforce Infrastructure

This is the people architecture behind AI transformation. What roles does the organization need to create? How do existing jobs change when AI handles more of the execution layer? Who is responsible for building capability across the workforce, continuously, not just at launch?

Workforce infrastructure covers role design, competency frameworks, learning systems, and performance management for human-agent teams. It is the pillar most squarely in HR's domain, and the one that most CoE frameworks treat as an afterthought.

2. Policy Infrastructure

Policy infrastructure is the set of frameworks, standards, and decision rights that govern how AI gets used across the organization. It operates at two levels. Enterprise-wide principles cover data security, ethics, compliance, and the boundaries that apply regardless of team or function. Team-level norms give people enough latitude to experiment and adapt AI tools to their specific workflows within those guardrails.

When policy infrastructure is missing, one of two things happens. Either the organization locks down AI use so tightly that adoption never gets off the ground, or there are no guardrails at all and the organization takes on risk it does not fully understand.

3. Strategy Infrastructure

Strategy infrastructure is what turns a collection of AI experiments into a coherent organizational capability. It answers the questions that determine whether AI investment actually delivers: what use cases are we prioritizing and why, how are we measuring real business outcomes rather than activity metrics, and what is our process for scaling what works?

Without strategy infrastructure, organizations end up with a long list of pilots that never go anywhere and no clear picture of what AI is actually delivering.

4. Integration Infrastructure

This is the technical layer: the tools, platforms, data architecture, and connection points that make AI usable inside the organization. It lives primarily in IT, but HR and L&D leaders have more influence here than most realize.

Tool accessibility is a direct driver of adoption. When employees have to navigate clunky platforms, switch between disconnected systems, or wait through lengthy approval cycles, adoption slows regardless of how strong the training and change management programs are. Getting involved in integration decisions early, particularly around user experience and access, is one of the highest-leverage things HR can do to support AI transformation.

5. Change Management Infrastructure

Change management infrastructure is the system that addresses the human dynamics determining whether people actually change how they work. Communications architecture, resistance frameworks, champions network design, and feedback loops that connect employee experience back to CoE decision-making.

The organizations that treat change management as a buildable organizational asset, rather than a communications plan they run once at launch, have dramatically better adoption outcomes. The AI Champions network is the most underinvested piece of this pillar. Champions are not trainers. They are trusted peers who normalize AI use, surface real friction, and create the conditions for adoption to spread organically.

What to do with an AI CoE framework

Now that you have the full picture, the most useful thing you can do is look at your organization against each pillar and ask honestly: what do we have, what are we missing, and who owns it?

For most HR and L&D leaders, that audit surfaces the same gap. Four pillars have a home somewhere in the organization, often in IT, legal, or strategy. The workforce pillar is either unowned, underdefined, or sitting on someone's to-do list with no clear path forward.

And that’s where HR and L&D can lead. Need help? Contact us to get started.