The MAO Framework: How CFOs Should Actually Think About Enterprise AI

There is a reason most enterprise AI investments underdeliver.

It is not the technology. The models are good. The vendors are capable. The budgets, in many cases, are substantial.

The problem is that most organizations treat AI as a procurement decision when it is actually an architectural one. They buy tools without designing systems. They automate tasks without rethinking workflows. They deploy models without building the operating infrastructure to sustain them.

The result is a portfolio of AI initiatives that each work in isolation and accomplish nothing at scale.

After years of designing AI systems at OpenAI, Spotify, Coinbase, and now through AndMaverick, I have watched this pattern repeat across industries, company sizes, and leadership teams. And I have come to believe the root cause is almost always the same: organizations do not have a coherent framework for thinking about what enterprise AI actually is.

The MAO Framework is mine. It is how I structure every engagement. And it is the clearest lens I know for a CFO trying to make sense of where the money should go.

Three Layers. Three Business Questions.

The MAO Framework organizes enterprise AI into three layers, each answering a distinct business question.

They are not independent components. They build on one another in sequence. That sequence matters.

Model: What Should the AI Think?

The Model layer is the intelligence layer.

It covers the selection, design, and governance of the AI systems that generate decisions for the business. Which models are appropriate for which problems. How they are trained, fine-tuned, or prompted. How their outputs are evaluated and trusted. How they are governed so the business remains in control of what the AI actually believes.

Most organizations start and stop here. They evaluate vendors, run pilots, pick a model, and declare an AI strategy. This is like hiring a brilliant analyst and then giving them no brief, no data access, and no authority to act.

Intelligence without execution is expensive decoration.

The CFO question at this layer: are we selecting and governing AI systems based on the decisions our business actually needs to make, or based on what vendors told us was impressive?

Action: What Should the AI Do?

The Action layer is the execution layer.

It is where intelligence gets embedded into the workflows, people, and software that actually create business value. This is where AI stops being a capability and starts being an operation. Where a model's output becomes a sales motion, a risk flag, a content decision, a customer interaction, a supply chain adjustment.

Most organizations underinvest here relative to the Model layer. They spend heavily on the AI itself and minimally on the integration work that makes it useful. The result is a powerful system that touches nothing consequential.

The Action layer requires two things most AI roadmaps do not address explicitly: process redesign and change management. You cannot embed intelligence into a workflow that was designed for humans operating without it. The workflow has to change. The people have to change with it.

The CFO question at this layer: are our AI investments actually changing how work gets done, or are they running parallel to the business without connecting to it?

Orchestration: How Does the Entire Organization Operate With AI?

The Orchestration layer is the operating layer.

It is the hardest to build and the most valuable when done. Orchestration coordinates models, systems, data, governance, and human oversight into a single enterprise AI capability. It is the difference between a collection of AI tools and an AI operating model.

This is where most organizations have nothing. They have vendors. They have pilots. They have a slide deck that says "AI Strategy" on the cover. What they do not have is a system that makes their AI investments compound on one another, that surfaces institutional intelligence rather than departmental noise, that gives leadership a coherent view of what the AI across the enterprise is actually doing and deciding.

Orchestration is what I designed at Spotify. It is what was missing at FTX when the crisis hit. It is what Coinbase had that kept it standing when every other exchange was collapsing.

It is not a technology purchase. It is an architectural commitment.

The CFO question at this layer: do we have a single operating model that coordinates our AI investments, or do we have a portfolio of unconnected experiments with no cumulative value?

Why the Sequence Is the Point

Model answers "What should the AI think?"

Action answers "What should the AI do?"

Orchestration answers "How does the entire organization operate with AI?"

This progression mirrors how businesses create value: intelligence, then execution, then operating model. A CFO understands this intuitively because it is how every other function in the business works. Finance has intelligence systems, execution processes, and an operating model. So does supply chain. So does sales.

AI is not different. It has just been sold as if it were.

The organizations that are pulling ahead right now are not the ones with the most advanced models. They are the ones that have built all three layers deliberately and in sequence. They started with the right intelligence. They embedded it into the right workflows. And they built the operating infrastructure to make it compound over time.

That is what an AI operating model looks like. That is what MAO is designed to produce.

The Question Every CFO Should Be Asking

Before the next AI budget cycle, before the next vendor presentation, before the next board update on AI progress, ask your team one question:

Which layer are we actually investing in?

If the answer is only Model, you have expensive intelligence with nowhere to go.

If the answer is Model and Action but not Orchestration, you have execution without compounding.

If the answer is all three, in sequence, with deliberate architecture connecting them, you have an AI operating model. And that is the only version of enterprise AI that creates durable competitive advantage.

Everything else is a pilot program with a good story and no second chapter.

Andrew Quillen is the founder of AndMaverick, a global Enterprise AI Orchestration consultancy. The MAO Framework is proprietary IP developed through engagements with enterprise and government clients across technology, finance, and media. To discuss how it applies to your organization, visit andmaverick.com.