How CFOs should be strategizing about generative A.I. spend

Good morning.

It’s been about nine months since OpenAI released ChatGPT and opened the world of generative A.I. to basically everyone. What began as apprehension for many companies—some even banning employees from using the technology—has turned into figuring out use cases to implement it.

CFOs understand how generative A.I. is changing business. And companies are preparing to spend big. I had a conversation with Michael Schrage, a research fellow at the MIT Sloan School Initiative on the Digital Economy, who says increased spending signifies companies are understanding that the technology matters.

However, "The issue isn’t ‘What should we spend?'" he says. "It’s ‘What do we want our returns to be?’” In his current work, Schrage is looking at how generative A.I. does, or does not, measurably contribute to key performance indicators and outcomes.

Use case scalability, the user journey, and risk-adjusted return must be high on the priority queue, he argues. For example, if someone says their generative A.I. spend has shot up from $500,000 to almost $3 million because of its “unique power and potential,” the appropriate response would be "What's your risk-adjusted expectations for returns in cost savings, speed to market, or top-line growth?” he says.

“If you truly understand and structure your use cases for generative A.I. correctly, there's much less risk associated with the investment,” Schrage says. Large language models are great at generating scenarios that let you calculate and calibrate the performance measures that matter, he argues. For example, “If you're trying to reduce the average length of a call or contact from two minutes to one minute, that’s doable,” Schrage says.

When thinking of KPIs, a question to keep in mind is: “How do we want to gauge/measure the returns?” Schrage says. “If I'm the CFO, do I defer to the ROI metric that IT or the call center leaders want? Or, do I have something to say about how they should really measure returns?”

Instead, a CFO might say, “Here's the metric that we're going to collaboratively design and agree upon to assure that you're getting adequate returns,” Schrage says.

Now, in terms of capital allocation, CFOs should be thinking: “What do I get for my investment in generative A.I. versus better analytics versus robotic process automation,” Schrage says. He notes that economists are always supposed to ask, “Compared to what?”

Best practices regarding generative A.I. will continue to evolve, according to Schrage.

“The only thing we can say for sure is that whatever the situation is in September 2023, is going to look different in January 2024,” he says. And he’s also sure that generative A.I. KPIs will be a collaborative effort, with CFOs talking more with CTOs, CIOs, and other C-suite members, he says.

Sheryl Estrada
sheryl.estrada@fortune.com

This story was originally featured on Fortune.com

More from Fortune:
5 side hustles where you may earn over $20,000 per year—all while working from home
Want more for your money? These 9 savings accounts have rates of 5.00% APY (and higher)
Buying a house? Here's how much to save
This is how much money you need to earn annually to comfortably buy a $600,000 home