Why CFOs — Not Chief AI Officers — Are the Secret to Getting Real Value From AI
Typically, value accountability for AI falls on the chief data and analytics officers or chief AI officers, Laks Srinivasan, co-founder and CEO of the Return on AI Institute, told me. But when CFOs oversee AI projects and are responsible for scoring outcomes, companies tend to extract more value, he said.
Srinivasan, an AI strategy expert, co-authored the study, “Economic Maturity for Artificial Intelligence,” with Thomas H. Davenport, a Babson College professor, MIT fellow, and co-founder of the Return on AI Institute. The findings are based on a survey of 1,006 C-suite executives across 11 countries and 32 industries, plus interviews with technology, data and AI leaders.
Only 2% of respondents said CFOs are charged with achieving value from AI. However, when CFOs are responsible, 76% achieved a great deal of value, substantially higher than for other roles. It’s not that CFOs necessarily know more about AI than a chief AI officer or other C-suite leaders, Srinivasan said. Finance chiefs can develop the methodology and scale it enterprise-wide. “When finance gets involved, it brings institutional credibility behind numbers,” he said.
In several companies surveyed, CFOs and finance teams partnered with technology executives to certify AI value. “For example, at DBS Bank in Singapore, the unit CFOs are responsible for vetting the AI value numbers before they are rolled up into the enterprise,” Srinivasan said. “And DBS Bank says it has generated about 1 billion Singapore dollars in economic value from its data analytics and AI initiatives; that’s because CFOs get involved,” he said.
The Return on AI Institute launched about five years ago and partners with Scaled Agile, Inc., on thought leadership and AI upskilling. Another key finding: generative AI is the most difficult type to establish value from, with 44% of respondents citing it, likely due to challenges measuring productivity for “broad and shallow” use cases.
Agentic AI ranks second at 24%, followed by analytical AI at 16%, while rule-based AI is the least difficult. Despite this, the 35% of companies that have adopted agentic AI report high value.
“From a personal, individual productivity perspective, I think we’re all seeing value,” Srinivasan said. Translating that to enterprise value is the challenge, he said.
His advice: involve finance. If teams track different metrics, aggregate them. “It may not be a science, maybe there’s a little bit of art involved, but you have to do it,” he said.
Another recommendation: AI upskilling for all. There’s a 23-point advantage in achieving high value when both employees and leaders are trained, yet 58% of organizations haven’t trained employees in basic AI use.
On workforce impact, only 2% of organizations surveyed have made large AI-driven headcount cuts, but nearly 90% have reduced or frozen hiring in anticipation. “Clearly, the headcount reductions and hiring freezes are running way ahead of evidence,” Srinivasan said. AI implementation also requires significant organizational change.
He recommends “narrow and deep AI” — reimagining specific processes for the AI era. Rather than layering AI onto existing workflows, the question becomes: what gets automated, and what still requires human judgment?
“You can actually make a solid, logical case to say, ‘This is really the headcount we need,’ after you do all the hard work,” Srinivasan said.
This article was written by Sheryl Estrada from Fortune and was legally licensed through the DiveMarketplace by Industry Dive. Please direct all licensing questions to legal@industrydive.com.
