How CFOs Can Drive Real Returns From AI Investments

It’s no secret enterprises are racing to deploy artificial intelligence across their organization. But corporate financial leaders are finding it’s often difficult to assess what AI actually costs, and what kind of return they’re getting on their investment. 

What’s more, this uncertainty is poised to increase as enterprise AI spending enters a second, more complex phase; one where costs don't end at initial pilots or licenses, but expand into governance, infrastructure, talent, security, and vendor sprawl. 

Faced with this challenge, CFOs must re-examine their approach to AI budgeting, oversight, and accountability to ensure they’re getting the most out of their investment in AI, and capturing the full potential of this transformative technology to optimize critical aspects of their business. 

The Hidden Costs of AI 

It’s no secret that corporate investment in enterprise AI is booming, projected at a whopping $2.5 trillion worldwide in 2026, according to Gartner. However, in a recent PwC survey of more than 4,450 business leaders across the globe, 56% reported they have yet to see any significant financial benefit from their adoption of AI. 

So what accounts for that gap between investment and return? 

According to Old National Bank Chief Information Officer Matt Keen, it’s not a failure of the technology itself to deliver value. Rather, it's because many companies have yet to develop models that accurately measure AI-driven costs. In particular, one aspect many organizations fail to price into their AI investments is the permanent operational infrastructure required to sustain these systems and tools, he said. 

“I think about it as ‘blue dollars’ versus ‘green dollars,’” said Keen. "Green dollars—money leaving the door—is easy to see: licenses, vendor contracts, tool purchases. But the blue dollars are where things get complicated: people's time, governance overhead, the opportunity cost of building and maintaining all of that infrastructure. That's where organizations often underestimate.”

Meanwhile, adding yet more unexpected costs to the equation, enterprise software vendors are embedding AI into existing products, often without full transparency. This means the actual number of AI touch points for an organization can go far beyond the AI systems and tools it has directly deployed, increasing not just direct costs, but also potential third-party risk. 

To improve its visibility into AI cost and risk exposure, Old National built a Center of Excellence to aide in vendor AI assessments, tracking which platforms embed AI, how those features are being used, and what cost and governance obligations they create. 

“You have to ask the right questions in your third-party risk management process, because you're still accountable for what's happening inside vendor platforms, even if you didn't build it,” Keen cautioned. “And it's not a one-time thing. You have to continuously monitor to keep up with any changes,” he added.

The Benefit Measurement Problem

Along with underestimating the costs associated with enterprise AI investment, many organizations also lack a clear picture of the other side of the ROI equation, and are unable to fully and accurately gauge the value of AI deployments. That’s largely because key metrics around AI-driven time and efficiency benefits remain somewhat novel and have yet to become firmly established and accepted throughout many companies.

But a closer look often reveals that AI is driving measurable improvements for a given use-case or application. For instance, Old National found a clear proof point around its own use of AI in the context of its call center, where AI-powered automation shortened average call handling time significantly, leading to meaningful savings for the bank, noted Keen. 

"That extra 30 seconds translates to X number of dollars, because we can measure what that efficiency is worth," Keen said. 

Commercial lending is another area where AI can create meaningful efficiency gains, Keen added. Tools that synthesize complex financial information can give underwriters a clearer view of the data they need to evaluate, helping teams move through analysis and decision-making more efficiently.

For any given use case, the key to assessing AI’s value lies in identifying and refining the right metrics, Keen stresses. In fact, AI itself can help surface those metrics by analyzing operational data to uncover patterns and opportunities for improvement that may not have been visible before. Keen noted examples of using AI to show process and efficiency improvements.

“It’s not always so much just taking an operational process and applying AI to it,” said Keen. “Firms can also use AI like an intelligent partner to help frame problems and figure out the right question to ask.”

Why Communication Is Key

According to Gartner, only 36% of CFOs say they feel confident driving enterprise AI impact. To help achieve organizational alignment and gain buy-in from key stakeholders on AI efforts, Keen recommends clear communication and transparency across finance and technology leaders. 

"It starts with budgeting; being able to articulate the needs, what the costs look like, and having those in a clean model,” Keen advised.

Transparency around ongoing costs is also vital, as AI outlays frequently appear as single line items inside invoices from vendors already under contract. 

"It’s important to be clear about all the different ways AI is coming into the organization," Keen said. "Leadership should have a clear, transparent picture of all the different places AI costs are actually showing up.”

To help ensure its own AI strategy is supported by organizational alignment and communication, Old National has established an AI Center of Excellence. The cross-functional initiative serves key functions including establishing acceptable use policies and maintaining a live inventory of AI across the organization, including within vendor platforms. Crucially, the CoE also acts as a structured forum for employees to bring ideas forward and move them toward production. 

"It's not just a governance and inventory model,” said Keen. “It's actually a way to promote and educate around potential AI uses as well, and provide a structured path from idea to implementation,” Keen said. “Without that clear pathway, organizational enthusiasm for initiatives can dissipate or get channeled inefficiently.”

AI as Strategic Partner 

As enterprise AI continues to radically transform a wide range of key business functions, and the pace of AI advancement ever-quickening, Keen advises organizations adopt an AI-first orientation—not just when it comes to operational functions, but strategic ones as well.

“I think one of the biggest transformations that leaders can bring is just thinking in a more broad way about how AI can help solve problems, ask interesting questions and figure out new and better strategies,” Keen noted. 

As an example of AI’s potential to inform and improve strategy, Keen suggested using AI systems to analyze large volumes of market, operational, or competitive information in ways that help organizations identify potential opportunities or areas where they can gain an advantage over peers.

"That's not expensive to do," Keen noted. "But it gives you very different insights than if you think you already have all the answers." 

In summary, Keen recommends treating AI as a uniquely valuable, highly skilled team member, and putting it to work on optimizing the most important critical aspects of your business.  

“If you had the most intelligent person on the planet in your office, you wouldn't send them off on a menial task. You'd ask them really interesting and important questions, said Keen. “That's the mindset shift that will separate the organizations that will become enterprise AI leaders.”

Key Takeaways: 

• Enterprise AI spending is accelerating, but many organizations struggle to accurately measure total costs and realized returns — especially as hidden operational, governance, and vendor-embedded AI expenses accumulate.
• CFOs can close the ROI gap by building clearer cost models, identifying measurable use-case metrics (like efficiency gains in call centers or underwriting), and continuously monitoring third-party AI risk exposure.
• Strong cross-functional communication, structured governance models like AI Centers of Excellence, and an AI-first strategic mindset can help finance leaders turn AI from an expense line into a measurable value driver.