"Surge Moment": Generative AI Disrupts Time-Tested ROI Metrics

It could be called the generative AI paradox.

Chief financial officers are chasing the potentially rich rewards of generative artificial intelligence, one of the most fervent waves of technology investment in decades.

Yet, according to finance executives and AI experts, they have at best a vague grasp of how to measure the ultimate business returns of this innovation.

"I have never seen anything like this," said Daniel Rock, who co-founded the generative AI consulting firm Workhelix with Erik Brynjolfsson and Andy McAfee.

Since the public release of ChatGPT in November 2022, "there was this 'aha!' moment where people said, 'It can do things that I didn't think information technology systems could do,'" Rock said in an interview. He noted that AI promises to streamline operations, better serve customers, upgrade forecasting, accelerate software programming, and improve marketing and sales.

San Francisco Federal Reserve Bank President Mary Daly said CFOs' embrace of generative AI is likely to be especially rapid because adoption is occurring largely bottom-up (from consumers) rather than top-down (from enterprises).

The release of ChatGPT marked a "surge moment" because "generative AI went overnight from something other people do to something that is relatively ubiquitous," Daly said at a March University of Chicago conference. "That moment brought what I call 'human-driven diffusion.'"

As a result, Daly said, CFOs and their executive peers face shareholder pressure and growing expectations for AI-driven efficiency and customer service. They must move "from indifference to action" faster than they did with other formative innovations such as steam power, electricity, and the internet.

"Our conclusion is that you will basically see the benefits of generative AI within the next five years," EY-Parthenon Chief Economist Gregory Daco said in an interview. Daco said it took computers 10 years, electricity more than 40 years, and steam power 80 years to achieve similar returns.

Rapid adoption of generative AI will enable companies to better innovate, leverage new skills of employees, retain customers, and gain market share, Daly said.

At the same time, according to finance executives and AI experts, rapid deployment of generative AI requires CFOs to measure ROI faster than with previous technologies, or it could result in wasteful spending.

The AI shakeout

CFOs must also be wary of hype and outright fraud, the top enforcement official at the U.S. Securities and Exchange Commission (SEC) said.

"While it may not yet be a perfect storm, there is certainly one brewing around AI," said Gurbir Grewal, director of the SEC's Division of Enforcement. "Every day, we see revolutionary technological advances in artificial intelligence (AI) that promise to transform nearly every aspect of our lives, including our financial decisions."

"Every day, we see individuals, companies, analysts, and others touting these advances," he said in an April speech. "Every day, we see companies not only trying to develop AI capabilities, or use it to enhance productivity and growth, but also trying to attract and retain investors."

According to a PwC survey, a majority of investors (61%) believe that rapid adoption of AI is very or extremely important to a company's value.

In turn, CFOs and their executive peers harbor a "fear of missing out," Glenn Hopper, CFO of Eventus Advisory Group, said in an interview. "They are under pressure from investors and boards—every moment, from everyone."

Consumers are also setting a high bar, finance executives and AI experts said.

"In some companies, AI deployment will become a must-do because customer expectations demand it," said Rock, who is also an assistant professor of operations, information, and decisions at the University of Pennsylvania's Wharton School. "It's just table stakes."

In the coming year, 37% of U.S. companies plan to invest at least $100 million in generative AI, a KPMG survey of 220 companies released in March found.

According to Gartner, AI software spending will surge to $298 billion by 2027, up 140% from 2022, with 35% of that investment focused on generative AI.

"You can't stop this momentum," Hopper said.

Nvidia, arguably the hottest momentum stock of the decade, leads the juggernaut behind generative AI adoption. Its stock price has soared 153% over the past year.

Yet, a hint of speculation about the operational value of Nvidia's products may signal that the influx into this new technology will enter unstable territory at some point.

Countries and companies alike are hiring the company to upgrade $1 trillion worth of data centers into "AI factories to produce a new commodity—artificial intelligence," Nvidia CEO Jensen Huang said on a May 22 quarterly earnings call. "The next industrial revolution has begun."

Nvidia said on a "company overview" page titled "High ROI for High Compute Performance" that cloud service providers can generate $5 in revenue over four years for every $1 invested in Nvidia software and networking.

However, in a footnote, Nvidia said in its May report that its calculation was merely an "illustrative example."

Admittedly, finance executives and AI experts said CFOs measuring the ROI of generative AI are trying to map out a murky terrain. Some returns from this emerging technology may be illusory, especially in the short term.

McKinsey, citing a global survey of 1,363 executives in May, said only 5.2% of organizations attribute more than 10% of their earnings before interest and taxes to the use of generative AI.

Big data, big challenges

The challenge of measuring the value of data predates generative AI and the emergence of so-called big data and advanced analytics two decades ago.

CFOs and investors have for years struggled to precisely assess the value of proprietary data about customers, markets, products, and other information.

The measurement challenge has not stopped the rise of Amazon, Google, Uber, and other Silicon Valley giants that rely on detailed, self-generated data to drive profits and growth.

These pioneering companies rode the explosive growth in the use of the internet, smartphones, and other connectivity and data-sharing tools. Laura Veldkamp, a finance professor at Columbia Business School, said at a February New York Fed conference on productivity that investors value their proprietary data more than buildings, workers, or technology.

CFOs in industries that existed before the advent of so-called information technology often face greater challenges in measuring the returns of generative AI in their operations, Rock said. "I'm not sure the total rate of return will exist."

Some companies' AI cost estimates are off by up to 1,000%, Gartner vice president and analyst Nisha Bhandare said at Gartner's May CFO conference. "Given how new AI is, we don't really know what it costs; we're learning as we go."

Editor's note: This is the first of two reports on the challenge of measuring returns on generative AI investments. In the second, CFO Dive will describe how finance executives are limiting costs and maximizing returns from the technology, even if they may initially lack reliable estimates of potential benefits.