CFOs face AI ROI gap

Finance teams are not arguing about whether AI matters anymore. That part is basically over. The live fight is much more CFO-shaped — where does the return show up, how fast, and can anyone prove it without breaking controls? That question got sharper this week as new survey coverage, finance-sector analysis, and accounting commentary all pointed at the same gap: spending is real, enthusiasm is real, but measurable ROI is still lagging. ### What changed this week? A PYMNTS report published May 8 said 34% of CFOs see increased output as the top reason to adopt AI. Behind that, 24% pointed to competitiveness and 19% to better decision-making. The survey focused on CFOs at companies with at least $1 billion in annual revenue, which matters because these are not tiny experimental budgets. (pymnts.com) ### Why is productivity the first target? Because productivity is the easiest promise to make. Finance leaders can picture faster close cycles, quicker reconciliations, less manual review, and fewer hours lost to repetitive work. Gartner’s February 2026 budget research lands in the same place — nearly 60% of CFOs planned to increase finance AI investment by 10% or more in 2026, and 88% ranked finance staff productivity among their top three priorities. (pymnts.com) ### So where does the ROI gap come from? The catch is that “useful” and “measurable” are not the same thing. Deloitte framed the core problem in March pretty cleanly: AI is now a core budget line, and CFOs are the ones who have to manage cost and measure return. But many teams are still stuck between pilot-stage wins and enterprise-grade proof. RGP’s 2026 CFO research report makes that gap blunt — only 14% of CFOs say they see clear, measurable impact from AI today, even though 66% expect significant positive ROI within two years. (gartner.com) ### Why is finance harder than a demo? Because finance runs on messy systems, not clean prompts. The Yahoo Finance piece on May 9 argued that the missing piece in banking is process intelligence — basically, a live map of how work, systems, and decisions actually connect. Without that, AI gets dropped into fragmented workflows and can’t reliably drive outcomes. It noted that less than half of banking AI projects are considered fully successful and meeting ROI expectations, while BCG estimates up to 60% of banking tech spend goes to maintaining existing systems. (deloitte.com) ### What does “without context” really mean? It means AI can generate output without understanding the business rules wrapped around that output. CPA Practice Advisor’s May 8 piece made the point in finance terms — if an AI tool cannot produce auditable, trusted results inside governance constraints, the productivity gain is fragile. The article’s answer was a “context graph,” but the broader point is simpler: finance AI has to know the chart of accounts, approval paths, tax logic, entity structure, and policy exceptions, not just the language of a task. (finance.yahoo.com) ### Is this also a people problem? Yes — and CFOs know it. PYMNTS’ broader workforce-readiness coverage says only 60% of firms feel even somewhat prepared for AI’s workforce impact, and just 12% say they are very prepared. At the same time, half of CFOs expect AI to create new roles requiring new skills, while 47% expect significant headcount reductions. That is a recipe for hesitation unless training and operating models catch up. (cpapracticeadvisor.com) ### What are firms doing about it? They are moving from “buy a tool” to “build a control system.” AICPA and CIMA launched an AI Accelerator Skills Program on April 29 aimed at governance and responsible deployment. Big firms are also pushing AI deeper into assurance and compliance workflows, like EY’s April rollout of agentic AI in assurance. The market is telling finance teams that adoption alone is no longer impressive — disciplined implementation is. (pymnts.com) ### Bottom line? CFOs do not have an AI interest problem. They have an evidence problem. In 2026, the winners will probably not be the teams with the flashiest copilots. They will be the ones that can tie AI to cycle-time gains, cost savings, cleaner controls, and decisions that hold up under audit. (rgp.com) (cpapracticeadvisor.com)