ROI or bust: CFOs pull plug on AI projects lacking results.
CFOs and boards demand AI ROI - now.
Corporate leaders are sounding a sharp new note on their AI investments: show us the ROI or expect budget cuts. A just-released survey of 260 finance executives finds that 87% say they must tie AI spending to business outcomes within the next year ([1]). The catch? Only 22% can connect any of their AI spend to actual business results today ([2]). In other words, nearly four out of five CFOs are flying blind on the returns from their AI projects, even as they’ve poured millions into pilots and tools.
Boardrooms are losing patience with this ROI vacuum. Where many directors once trusted the AI hype, now two-thirds of corporate boards insist on seeing evidence of value before approving any new AI funding ([3]). In fact, 22% of boards won’t greenlight a single additional AI initiative unless ROI is demonstrated upfront ([4]). This is a startling shift from the “blank cheque” era of the past two years when AI budgets often escaped normal scrutiny. Executives say the blank-cheque days are over ([5]), replaced by rigourous ROI reviews for every project. CFOs are under pressure to answer the question that used to be an afterthought: “What are we getting for all this AI spend?”
Halting projects and the new cost calculus.
With scrutiny at an all-time high, many companies are slamming the brakes on AI projects that can’t justify themselves in hard numbers. According to the finance leader survey, fully 75% of CFOs who lack clear ROI metrics have already put further AI investments on hold ([1]). More than one in three have gone further and cancelled at least one AI initiative outright, cutting their losses rather than throwing good money after bad ([2]). This pullback is not hypothetical - it’s happening across industries as CEOs and boards question experimental projects that failed to translate into measurable cost savings or revenue gains.
In a twist, “spending less” on AI is no safe harbour either. New data shows that companies coming in under budget often face even tougher questions from the board than those overspending. Teams that undershot their AI budget encountered board scrutiny 58% of the time - more than the 48% for teams that overspent ([3]). As one analysis put it, underspending signals you might be falling behind, while overspending looks reckless ([4]). Either way, directors are no longer asking how much was spent, but whether it delivered results. The message to management is clear: purely “activity-based” AI projects - whether over-zealous or half-hearted - will be cut unless they can prove their worth.
Shifting from hype to measurable value.
This tightening of the purse strings is forcing a profound change in how enterprises approach AI. Leaders who want to keep their AI budgets intact are realising they need to re-focus on metrics that matter to the business. A common mistake has been the “AI consumption trap” - measuring easy stats like API calls or tokens used and mistaking them for value ([1]). Gregg Aldana, a global AI advisor, notes that counting how many models or algorithms are deployed means little “without knowing what process changed” or improved as a result ([2]). In his experience, asking “What did we actually get for this AI spend?” often leads to uncomfortable silence. That’s why he urges executives to start with a specific workflow and define what success looks like - whether it’s faster claim processing in insurance, fewer stockouts in retail, or improved client conversion in banking. Only by linking AI to concrete operational improvements can companies identify genuine ROI.
An AI initiative “does not have an ROI in isolation” - the returns only materialise when AI is embedded into business processes that generate value ([3]). For example, spending an extra $1 million on AI to save $10 million in claims handling costs is a clear win, whereas a cheap AI tool that delivers no meaningful improvement is still a loss-making proposition ([4]). Many firms are now adopting this mindset. They are limiting AI programs to a handful of high-impact use cases, measuring baseline performance and tracking how AI enhances key metrics like cycle time, cost per transaction, error rates or customer satisfaction. As Aldana puts it, companies are learning to “connect AI spend to the economics of the work it’s supposed to improve,” instead of treating AI itself as the outcome ([5]).
Building an AI business case that delivers.
What distinguishes the organisations actually reaping AI rewards? Multiple studies in 2026 have observed that the top tier of AI “ROI leaders” take a very different approach than the rest. These leading companies obsess over aligning AI projects with business strategy and financial goals from day one ([1]) ([2]). They don’t chase dozens of experimental pilots; instead, they double down on a few well-chosen applications where AI can either boost revenue or cut costs in a measurable way. Crucially, they redesign workflows around those AI capabilities (making employees 2× as likely to embrace them) and invest heavily in change management and training. The payoff is clear: according to PwC and McKinsey data, roughly 20% of companies now capture the vast majority of AI’s economic benefits ([3]) ([4]), while the bottom 80% of firms are left fighting for scraps.
As boards and investors demand substance over buzzwords, tech leaders must treat AI as a business transformation lever, not a shiny toy. That means involving CFOs and COOs in AI governance, insisting on real-time cost visibility, and defining success metrics up front. Companies that built these disciplines were able to invest **more** in AI with confidence, even during budget cuts, because they could clearly demonstrate value creation from their deployments ([5]) ([6]). In contrast, those who failed to measure ROI are seeing funding dry up and find themselves stuck in “pilot purgatory.” The window for AI experimentation without accountability has closed. Now is the time for executives to take charge of the AI business case - or risk being left behind by more disciplined competitors.