CFOs urged to treat AI as a portfolio, not one ROI

Chief financial officers should rethink how they measure returns on artificial intelligence investments, according to Gartner. Many finance leaders treat AI as a single ROI question, when it should be assessed as a portfolio of different bets.

The case rests on a simple point: AI projects do not share the same cost base, timetable or value profile. Finance teams should separate routine productivity uses, targeted process improvements and larger transformational projects instead of forcing them through a single investment test.

Twisha Sharma, senior principal, research, in the Gartner Finance practice, outlined that view to finance executives in Sydney. She said AI economics vary sharply by use case, leaving companies exposed if they rely on a standard method to judge value.

"AI does not follow one cost curve, and it does not produce one uniform type of value," Sharma said. "CFOs need to stop looking for a single ROI formula and instead build a balanced portfolio that includes productivity use cases, targeted process improvements, and selective transformational bets."

Different models

The framework divides AI initiatives into three broad groups: routine applications that automate repetitive tasks, more advanced uses that improve analysis and decision-making, and larger projects aimed at innovation or market disruption.

For finance chiefs, the distinction matters because each type of initiative carries a different mix of spending, risk and expected return. Timelines also vary. Some projects produce measurable operational gains quickly, while others take longer before any commercial effect becomes visible.

Sharma said cost structures can differ considerably, making close scrutiny of spending assumptions essential. Weak cost analysis at the start of a project, she warned, can create budget problems later.

"The economics of AI differ sharply from one use case to another, making it difficult for a standard value approach to capture the full picture, especially as the cost difference between various types can be significant," Sharma said. "Each use case will have different timelines, different ambitions, different risk profiles and different ongoing costs. If finance teams don't dissect cost models with precision, they will face budget surprises later."

Beyond profit

Gartner also argues that a narrow focus on near-term financial outcomes can lead companies to undervalue AI. Some initiatives generate nonfinancial benefits first, before any effect appears clearly in revenue, costs or cash flow.

These benefits can include stronger decision support, faster organisational adaptation, broader reach across the business and greater scope for innovation. AI tools may also change finance's role inside companies by reshaping how teams support operating decisions and planning.

That matters for CFOs because traditional investment disciplines often place the greatest weight on direct, measurable financial returns. If AI projects are judged too early or too narrowly, businesses may cut initiatives that are strengthening processes or decision-making in ways not yet reflected in earnings.

Sharma said these softer gains should still form part of any serious assessment of AI performance. She described them as early signs of value creation rather than secondary effects.

"The value of AI is not always captured first in traditional financial metrics. In many cases, it appears earlier in better decisions, faster adaptation and stronger organizational capability. CFOs need to account for that if they want a complete picture of what AI is really delivering," Sharma said.

Portfolio approach

Gartner's broader message is that finance leaders should manage AI spending more like a portfolio than a single programme. That means backing a mix of smaller and larger projects, expanding those that show results and shutting down weaker efforts before costs rise further.

The approach reflects a more cautious stance on AI investment at a time when many companies remain under pressure to show clear returns from new technology spending. It suggests CFOs should focus less on finding a universal formula for AI and more on matching funding models and expectations to the use case under review.

For finance teams, the practical challenge will be building governance systems that can compare very different kinds of AI work without flattening the differences between them. Gartner's warning is that companies that fail to do so may misunderstand both the risks and the benefits of the technology.