AI is forcing finance to rethink one of its oldest assumptions: that you can measure the value of an investment with a single business case. That approach has worked for decades. It may not work for AI.
On The CFO Playbook, Riccardo Calliano, who runs finance for GSK’s GenAI commercial investments, argues that the way traditional finance measures return does not work for AI because it breaks the single calculation almost every finance team builds its spreadsheets around.
The assumption behind every business case
If you buy something, you weigh the cost against the savings, and the ratio tells you whether to go ahead. It is a payback period, clean and defensible, the same test you applied to the photocopier, the ERP migration and countless other investments. For years, this analysis has been the spine of every business case.
The assumption is that whatever you bought is finished the day you buy it. It does the same job now as it will do in two years, so a single snapshot tells you everything you need to know.
AI is different.
AI systems improve through new models, better data and the way people learn to use them. That means the capability finance approved six months ago is rarely the same capability it is evaluating today.
If you measure something that is still changing, the results can be misleading. If you assess an AI investment in the first few quarters, before the technology has matured and people have adapted how they work, the return may not reflect the value it can create over time.
AI also creates a visibility challenge. Finance is no longer evaluating a fixed asset. Unlike traditional software, it is often adopted gradually across teams, with value built through learning and how hard your people work with it.
Understanding who is using AI, how often and whether it is changing the way people work becomes just as important as understanding what was purchased.
Adoption is where value emerges
Monica Proothi, who leads Global Finance Transformation at IBM Consulting, argues that the hardest part of AI was never the technology. She describes this as stickiness: “Getting AI to stick with humans… to augment their day-to-day life”, the thing that separates investments that pay off from those that do not.
Stickiness is neither luck nor the vendor’s problem to solve, says Monica. Businesses need to get their data in order, put governance in place around AI and invest in training. AI should feel like the easiest way to complete a task, not another thing to figure out.
Both Monica and Riccardo believe that value emerges once the investment is embedded. That makes AI a human challenge as much as a technical one.
For finance, it is just as important to understand why nobody is using AI as it is to understand what it is costing the business.
This reflects a wider shift in finance’s role. As Riccardo puts it, “CFOs and finance organisations need to start to look at themselves not just as guardians of value, but drivers of transformation and innovation.” That means looking beyond what has already been spent to understand where adoption is slowing, and helping the business address it before value is lost.
That challenge is reflected in Soldo’s Productivity at Work research: while employees recognise AI’s potential, 58% say AI tools are difficult to use and 57% do not understand how AI is meant to help them improve their work. Adoption is not automatic; organisations need the right processes and visibility to turn technology into better decisions.
Governance still matters
None of this means finance should relax its controls. But the control finance needs now looks different from the control it’s used to.
AI spending rarely arrives through a single enterprise contract. It often appears gradually across departments as subscriptions purchased on company cards, making it difficult for finance to answer a simple question: what are we actually spending, and is anyone using it?
Governance still matters. But the focus is changing. Finance needs to understand where AI is being adopted before month-end, not simply whether spending stayed within budget.
Where’s the revenue?
The question many CFOs are asking is simple: if we’re spending more on AI, why is revenue not increasing?
Riccardo argues that the technology is rarely the problem. AI can create capacity. Whether that capacity becomes growth depends on the decisions finance helps the business make next.
For finance leaders, proving AI works is not just about measuring what was spent. It is about understanding whether the organisation is adopting it effectively and what needs to change to turn investment into impact.
