How CFOs should think about compensation and AI usage, according to rewards expert

We’re in a bit of a choose-your-own-adventure moment in finance when it comes to AI usage and employee compensation.

You’ve got the JPMorgan Chases of the world, with internal dashboards tracking AI usage, and then you’ve got EY doling out $100 million in bonuses for “human” skills. And then, presumably, you’ve got your company, which may land somewhere in between: Trying to encourage innovation and efficiency, but still focused on the core nontech ingredients that can really make a company thrive.

So, what now?

Before companies can consider redesigning pay structures for an AI-enabled workforce, Tauseef Rahman, a US workforce solutions leader at Marsh, recommends CFOs work through a much more fundamental question: Do we really understand which tasks are driving value?

“You can’t just paint technology onto broken processes and unclear value creation, and then be surprised it didn’t make things better,” he warned. If companies don’t understand a given task or role’s value, they also can’t tell what AI will do that role: Will it enhance it, hurt it, or simply change how it’s completed?

“Do we know why we are doing the work the way we are doing it?” Rahman recommended CFOs ask themselves. “How do we know that directing more budget into this area will produce the same, if not better, investments?”

Team spirit. The issue is landing on the CFO’s desk in part because someone has to own it. A recent Marsh survey found the percentage of employees who felt “no one owns” digital agility in their organization grew from near zero three years ago to 12% in 2026.

But it’s going to be a team effort, Rahman noted. To get started, you’ll want to “start with functions where you can more easily trace the outcomes created by an organization,” he explained, using a sales team as an example. “What is it about the sales teams and how they do the work, and the people that we have, that generates the sales that we do?”

To answer those questions, you’ll need to work across the organization.

“Sales, HR, and the finance function need to actually work through all that stuff together,” he explained. “The sales team knows how the work is done, how the sales are made. The finance team understands the impact of the net bottom line of the business in terms of the revenue, in terms of margin, cost of sales. Then, of course, the people team is supremely involved in understanding the job design and how that work needs to change.”

“Start where the sales are made or the revenue is generated, [and] just trace through ‘What is the org design and structure in terms of how that is produced, and are there better uses of people to do the work?’” Rahman recommended.

The “original sin” of AI at work. In some cases, the process might involve reconsidering previous tech investments that aren’t panning out. That may be particularly true for any organizations that unveiled compensation metrics around tech usage without considering if the actual outcomes of a given task had improved, Rahman said.

“It sounds like when computers came, we would measure keystrokes per minute on the keyboard, but that’s what some organizations do because they didn’t know how to understand value creation,” he explained. “You need to really be able to understand: Does the addition of technology into the flow of work result in better or greater output?”

“Organizations made the mistake of rewarding use, not outcomes,” he continued. “People assumed that tech usage inherently meant better outcomes. That’s the original sin in what’s happened with AI at work.”

Through it all, there’s a simple formula organizations should keep top of mind, Rahman said, and if there’s one thing CFOs love, it’s a good formula: “Outcomes over outputs, outputs over usage.”