I recently met with a client’s CEO and CIO together. The CIO asked, “How much should we spend on AI next year?” My response was, “If you do your job right, your CEO will ask why you aren’t spending more.”

I spend a lot of time with CIOs who are planning their AI investments. One pattern keeps repeating itself. Our discussion starts with token costs, model pricing and consumption forecasts. It rarely ends there.

Somewhere along the way, we start talking about which AI investments drive business performance (revenue, margin, customer satisfaction, churn, time to market, scale).  This turns AI budgeting from an IT cost to a business ROI.  I find this interesting because it reflects a fundamental shift for the CIO. That changes the CIO’s job from controlling AI spend to figuring out where the business should invest more and benefit more.

Consumption cost concerns are valid

For decades, technology leaders have worked with relatively predictable costs: buying servers and depreciating them or purchasing X software seats for Y dollars and renewing annually.

Consumption-based AI changes that equation. It’s more like paying a utility bill than owning an asset, and traditional IT budgeting wasn’t built for that variability.

Few CIOs I talk with have faced cost becoming a problem because adoption worked too well. We see some organizations considering constraining deployments or renegotiating contracts as costs increase. That’s the wrong strategic path. They’re managing costs when they should be more discriminating about which AI initiatives deserve investment.

Think like a capital allocator, and focus on value realization

Organizations realizing real value from AI are leveraging disciplines long used in finance and corporate strategy to optimize their capital investment portfolios. They are rigorous about evaluating the impacts that investments have on the business. There may be hundreds of projects underway, but only a handful matter most to the business right now. That’s where the resources go.

I’ve also seen this firsthand at West Monroe. Our adoption curve accelerated as people moved from experimentation to transforming work with AI-enabled workflows. We found that our top two percent of users accounted for roughly half of AI usage. They also had outsized impact across sales, delivery and other operations. That visibility shifted the conversation from, “Why are we consuming so much AI?” to “Where should we invest even more?” 

A recent IBM study found organizations that manage AI as a portfolio rather than a series of individual investments deployed 2.4 times as many agents without a higher AI/IT budget.

The economics of AI require a new management approach that connects AI activity to business economics and, subsequently, to investment decisions. Measurement, funding and governance must all change together, along with the CIO’s focus.

Measurement should prove what AI is producing

Few organizations can answer the question, “What did we get for what we paid?” Most measure tokens or similar usage rates. These are meaningful leading indicators because they reflect the foundation for value capture. But they aren’t measures of business value.

When evaluating value, many executives mistakenly equate correlation with causation: “We have a lot of AI use in a particular group. Revenue is up. That must be because of AI.” What they need to do is link AI activities to the KPIs that drive the business and ultimately impact revenue or margin. This is much harder because it requires a deep understanding of business processes. If you use AI to remove 10 steps from customer onboarding and, as a result, customers complete the process 70% of the time instead of 50%, you’ve achieved a 20-point increase in retention. You can then calculate how that translates to the P&L. 

Measuring value requires a system to model business processes before and after AI, and to understand how they drive KPIs and ultimately the P&L. The good news is that with AI, this type of modeling is possible. And new ways to measure value are fast emerging, such as return on tokens.

Modeling isn’t easy, so start small. Identify and measure one or two business metrics that AI could directly influence, then build the measurement framework before building anything else. Here’s an example: My company introduced training on using AI in the sales process and measured sales performance before and after. People who completed the training had a 47% larger pipeline and 57% more wins than those who didn’t. This insight helped us decide whether to accelerate our AI investment.

Funding should follow the value

In the early stages, it makes sense for IT to fund AI. Over time, though, keeping AI costs within IT limits flexibility because it competes for a finite budget, while other parts of the business reap the benefits.

We found that AI usage on delivery work for a key project cost us $18,000 per month but generated $500,000 in productivity gains, a 20X ROI. The team that does this work should obviously pay for the AI. If we left that cost in the IT function, our CIO would say, “You’re spending too much on AI; try to limit it.” That would reduce ROI.

By creating visibility into how AI activity drives business outcomes, CIOs can begin to shift costs to the respective function, as with labor or office space. In fact, if you treat AI as part of workforce costs rather than an IT cost, it becomes part of enterprise-wide operating expenses, rather than sitting entirely within the IT budget.

This isn’t easy or quick. There are always organizational politics, but this isn’t any different. Many CFOs don’t feel they have control over AI activity and its impact on the business. A CFO can’t reasonably fund what the organization can’t model. This goes back to the point above. Our organization built the modeling capability first. When other parts of the organization can measure it, price it into deals and plan for it, we move the cost out of IT. This insight also gives our CFO and business leaders more confidence in future investment decisions.

Governance must facilitate rapid adaptation

Traditionally, governance exists to reduce risk in a relatively stable environment. With AI, we’re governing through change. And, in this case, the change is continuous.  Everything — technology, cost structures and best practices — is changing rapidly. Organizations need an adaptive governance approach. Governance teams should be multidisciplinary, including business stakeholders, finance, risk and IT. They should meet frequently to understand and respond to a changing reality. Every month, they should ask: Are the things we planned to do in AI still the right ones? Will they have enough impact? Are we going fast enough? Are we being too risk-averse or taking too much risk?

A good analogy is gasoline consumption amid rising prices. If gas were free, we’d floor it all the time. But we’ve learned to manage consumption through various mechanisms, such as dashboard metrics and features that automatically stop and start the engine at traffic lights. Annoying as these are, we’ve become more efficient without driving less. That’s what’s happening with AI.

You can manage costs by suppressing consumption, or you can manage the economics by improving the return on consumption. The latter is what a portfolio manager does. That brings me to the CIO’s role.

IT must lead an intelligence transformation

I think of CIOs in eras. In the data center era, those who excelled at building data centers became IT leaders. Next came the SaaS application era. Now, we’re in the intelligence era. AI is most effective when it has access to data, which forces CIOs to consider how to make data more useful, effective and accessible, and what agents can automate. That’s new and, frankly, terrifying for many. It is easier to say “no” than to unleash agents.

I believe the CIO’s job has never been more challenging. AI is harder to budget for and forecast. There’s more pressure than ever to align technology spending with business outcomes and demonstrate ROI. But there’s also more opportunity than ever to do so.

Today’s most effective CIOs are closely aligning with the business and leading an intelligence transformation. Those perceived as gatekeepers will be replaced by more progressive individuals.

The CIO’s opportunity

The best AI budget conversations still begin with technology, as they should. But they quickly shift to where the organization should spend more, as in my recent conversation with the CIO and CEO. The CIO’s influence grows as AI spending becomes inseparable from business performance. Ultimately, this moves the role toward what it has long sought to be: a trusted partner to the business.