The Real AI Budget Question: Where Should We Spend More?

From Cost Center to Value Driver

The conversation around artificial intelligence budgets has fundamentally shifted. For years, technology leaders planned spending based on predictable models: buying servers, depreciating hardware, or purchasing software seats for a set annual fee. These were manageable, static costs that fit neatly into traditional IT budgeting frameworks.

That era is over. With the rise of large language models and AI agents, the economics of technology have changed. Spending is now consumption-based, resembling a utility bill rather than an asset purchase. This variability breaks old budgeting habits, forcing Chief Information Officers (CIOs) and business leaders to rethink how they allocate capital.

Refer & earn: 30% of a client’s first month, then 10% of every month after that, for up to 24 months per client. See the partner programme →

A recent discussion between a CIO and CEO highlighted this tension. When asked how much should be spent on AI next year, the response was stark: if the CIO does their job right, the CEO will ask why they aren’t spending more. This reversal signals a critical transition. AI is no longer just an IT expense to be controlled; it is a business investment to be optimized.

The Trap of Token Counting

Most organizations start their AI budgeting discussions with technical metrics: token costs, model pricing, and consumption forecasts. While these are necessary leading indicators, they are not measures of business value. Measuring only tokens tells you how much foundation work is being done, but it does not tell you what return that work generates.

Many executives mistakenly equate correlation with causation. They observe high AI usage in a specific group and see revenue increase, assuming the two are directly linked. However, without linking AI activities to key performance indicators (KPIs) that drive the business, such as churn, time to market, or customer satisfaction, organizations cannot prove ROI. This gap creates uncertainty for CFOs, who cannot reasonably fund what the organization cannot model.

The question “What did we get for what we paid?” remains unanswered for most companies. To fix this, measurement systems must model business processes before and after AI implementation. For example, if AI removes ten steps from customer onboarding and increases completion rates from 50% to 70%, that twenty-point increase in retention can be translated directly into profit and loss statements.

Treating AI as a Portfolio, Not a Project

Organizations realizing real value from AI are applying disciplines long used in finance and corporate strategy. They treat AI initiatives as a capital investment portfolio rather than a series of individual projects. In any given quarter, hundreds of projects may be underway, but only a handful matter most to the business. Resources are directed toward those high-impact areas.

Data from West Monroe illustrates this point. Their adoption curve accelerated as employees moved from experimentation to transforming work with AI-enabled workflows. The top two percent of users accounted for roughly half of all AI usage and had an outsized impact on sales, delivery, and operations. This visibility shifted internal conversations from concern about consumption levels to questions about where to invest even more.

An IBM study supports this portfolio approach, finding that organizations managing AI as a portfolio deployed 2.4 times as many agents without increasing their overall AI or IT budget. By being rigorous about evaluating impacts, companies can optimize their investment mix effectively.

Shifting Costs to Business Functions

In the early stages of AI adoption, it makes sense for IT to fund these experiments. However, keeping AI costs within IT limits flexibility over time. IT budgets are finite, and when other parts of the business reap the benefits of AI-driven productivity, funding them solely through IT creates misalignment.

Consider a scenario where AI usage on delivery work costs $18,000 per month but generates $500,000 in productivity gains, resulting in a 20X return on investment. If the cost remains in the IT function, the CIO may feel pressured to limit usage to control spend, thereby reducing potential ROI. Instead, the team benefiting from the work should pay for the AI. This mirrors how labor or office space costs are handled.

Treating AI as part of workforce costs rather than an IT cost allows it to become part of enterprise-wide operating expenses. This shift requires visibility into how AI activity drives business outcomes. Once other parts of the organization can measure and price AI into their deals, the cost moves out of IT. This gives CFOs and business leaders greater confidence in future investment decisions.

Governing Through Continuous Change

Traditional governance exists to reduce risk in stable environments. AI, however, involves continuous change. Technology, cost structures, and best practices are evolving rapidly. Governance teams must therefore be multidisciplinary, including business stakeholders, finance, risk, and IT. They need to meet frequently to assess whether current plans remain valid.

Monthly reviews should ask critical questions: Are the things planned still the right ones? Will they have enough impact? Are we moving fast enough? Are we being too risk-averse or taking too much risk?

A useful analogy is gasoline consumption amid rising prices. If gas were free, drivers would floor it constantly. But as prices rise, people manage consumption through dashboard metrics and features that automatically stop engines at lights. These mechanisms make driving more efficient without necessarily reducing distance traveled. Similarly, organizations can manage AI economics by improving the return on consumption rather than suppressing it.

The Evolving Role of the CIO

The role of the CIO has evolved through distinct eras. In the data center era, leaders excelled at building infrastructure. In the SaaS application era, they managed software subscriptions. Now, in the intelligence era, CIOs must ensure AI has access to useful, effective, and accessible data while determining which agents can automate tasks.

This new responsibility is challenging. AI is harder to budget for and forecast than previous technologies. There is immense pressure to align spending with business outcomes. However, there is also unprecedented opportunity. CIOs who closely align with the business and lead intelligence transformation will see their influence grow. Those perceived merely as gatekeepers risk being replaced by more progressive individuals.

The best AI budget conversations begin with technology but quickly shift to business performance. As spending becomes inseparable from revenue and margin, the CIO’s role moves toward that of a trusted business partner. The focus is no longer on controlling spend, but on figuring out where the business should invest more to benefit most.

Practical Steps for Leaders

For businesses deciding what to do about AI, the path forward involves several concrete actions:

• Start Small with Modeling: Identify one or two business metrics that AI could directly influence. Build the measurement framework before building the technology. Measure performance before and after interventions to establish causality.
• Link to KPIs: Move beyond token counts. Connect AI activities to metrics that drive revenue or margin, such as customer retention rates or sales pipeline size.
• Reallocate Funding: As AI proves its value in specific departments, shift the cost burden from IT to those business units. This ensures that those benefiting from efficiency gains contribute to the investment.
• Adopt Adaptive Governance: Create cross-functional teams to review AI initiatives monthly. Focus on optimizing return on consumption rather than simply limiting usage.

By treating AI as a strategic portfolio and rigorously measuring its impact on business economics, organizations can move from uncertainty to confident investment. The goal is not just to use AI, but to use it where it matters most.

Sources

The real AI budget question: Where should we spend more? – CIO