According to McKinsey research published last year, even top-performing companies capture only about 70 percent of their strategies’ full potential—often because their operating models can’t keep up. Which raises two practical questions: what is an operating model, and what happens to it when AI enters the picture?

Put simply, an operating model is how work actually gets done in your organization: how you deliver value, how decisions get made, how resources get allocated, and how strategy turns into day-to-day execution.

However, “operating model” is also one of those terms that gets thrown around a lot. Most leaders have an intuitive sense of what it means, but if you asked them to define it in one sentence, they’d probably struggle. That’s partly because an operating model isn’t one thing—it’s a system of choices that determines how work gets done.

McKinsey’s “Organize to Value” framing is useful for that reason. It breaks an operating model into 12 interconnected elements: purpose, value agenda, structure, ecosystem, leadership, governance, processes, technology, behaviors, rewards, footprint, and talent. These elements interact as a single system, and that system is your operating model.

Once you can see the component parts, you can then start asking where AI changes the assumptions your operating model is built on. In that sense, AI can become the catalyst for an operating model redesign.

The AI Operating Model Diagnostic

The primary question you should be asking about your organization is this:

Where will AI break our current operating model first, and what needs to be redesigned before we scale?

To answer that, you need to diagnose the operating model as a system. Break it into its 12 elements, then use the questions below to identify where AI will create friction first—and which redesigns will unlock scale.

1. Purpose: What should AI make meaningfully easier for customers and frontline teams—specifically?

2. Value agenda: Where will AI create measurable value first, and what are we explicitly not doing yet?

3. Structure: Where should AI capability live (centralized, embedded, hybrid), and who owns outcomes?

4. Ecosystem: What should come from partners/vendors versus being built in-house—and why?

5. Leadership: Which decisions should leaders make differently as AI changes the speed and quality of information?

6. Governance: What are the rules for data, risk, approvals, and human review?

7. Processes: Which workflows should be redesigned end-to-end rather than “AI bolted on,” and what becomes the new standard?

8. Technology: What is the approved AI stack and integration path, and how do we ensure permissioning and auditability?

9. Behaviors: What new behaviors do we expect (sharing, documenting, QA), and what becomes unacceptable (shadow tools, unreviewed outputs)?

10. Rewards: Are we rewarding measurable impact, or just activity and experimentation theater?

11. Footprint: Where does AI change where work happens (central vs field, shared services vs business units), and what does that imply about roles?

12. Talent: What capabilities must be built by role—and how will we train and reinforce them over time?

You don’t need perfect answers to all twelve to start. In fact, you’ll probably struggle to answer a few of them. The point here is to find the weak links, because an operating model behaves like a system: one constraint can limit everything above it.

Going back to the McKinsey piece, the article describes the outcomes of an effective operating model in four words—clarity, speed, skills, and commitment. That’s a useful scoreboard here too.

If AI is going to change how work gets done, those are the four outcomes you should expect to improve. If they aren’t improving, the issue usually isn’t the AI model—it’s the operating model underneath it.