Your AI bill tells you what you spent, but not whether anything got better.
That divide worries finance leaders. Mike DePrisco, President and CEO of the Institute of Management Accountants (IMA), speaks with hundreds of CFOs and finance leaders around the world. Their questions have evolved from whether to use AI to how to get value from it.
On the Earmark Podcast, Mike and host Blake Oliver, CPA, explored how accountants can measure that value. The challenge is finding business improvements and developing people who can check, interpret, and explain AI’s output.
Adoption isn’t the same as a return
AI is the top concern among the finance leaders Mike meets. They’re deploying and investing in it, but skepticism is growing about the payoff.
“We’re not seeing the promised return on investment fast enough,” he says, describing what he hears from leaders.
Blake points to a related problem: rising costs. He describes a shift from predictable subscriptions of roughly $20 to $100 per seat toward additional usage-based charges from providers like Anthropic and OpenAI. Those charges are measured in tokens, the units used to track model usage, and CFOs and controllers struggle to keep spending under control.
Mike says accounting and finance professionals adopted AI slowly, partly because mistakes can affect financial statements, reporting, and compliance. That caution was reasonable. Now that more organizations use the technology, they need tangible evidence that it works.
But tracking spending is only half the problem. What should leaders measure on the other side?
Measure what actually improves
Blake notes that token costs are measurable. Connecting those costs to outcomes is harder. Mike offers several practical indicators, not as a complete ROI formula, but as ways to judge whether an investment produces value.
- Finance operations. “Are you closing the books more quickly?” Mike asks. Are you reducing errors month over month? Together, these measures help show whether AI improves speed and quality.
- Decision support. Is forecasting more accurate? Mike connects better forecasts with better decisions and better use of resources. More output alone doesn’t prove value; the information needs to help the business make decisions.
- Customer experience. Does the technology address actual customer problems? Internal customers may need to get work done faster or at lower cost. External customers may want smoother transactions or easier searches.
For financial institutions, Mike gives a specific example of reducing false fraud flags that put customers’ accounts on hold. That’s a business outcome, not just an improvement inside the finance department.
Organizations IMA works with are starting to use these measures to decide whether AI is working and whether to continue, scale, or invest more.
Mike suggests the strongest results come from having a clear problem to solve. But even a useful application still needs people who can question its output.
Faster work still needs sound judgment
Blake raises concerns about AI reducing entry-level work in public accounting. Mike isn’t fully convinced that the technology will mean fewer jobs overall. He expects jobs to change, but sees a continuing need for accounting knowledge.
“We see the mistakes that AI makes every single day,” he says. People with business knowledge and customer context spend time correcting those mistakes.
He describes this as human-machine collaboration. People must stay involved to check for bias, verify accuracy, and interpret information before using it to make decisions.
That’s why foundational training is still important. When Mike asks faculty how they respond to machines taking over basic work, they tell him students still need to understand the fundamentals. If students are going to work alongside these machines, “they need to know what to look for.”
The role also extends beyond checking numbers. Mike describes management accountants as business partners who interpret data from AI, ERP systems, and forecasting tools. They help other functions assess opportunities, solve problems, and make decisions.
That creates a training challenge. How do accountants build judgment if technology takes over the routine work where earlier generations learned?
Build the skills AI can’t supply on its own
IMA expanded its competency framework to include advanced data analytics, transformation, innovation, sustainability, and leadership. Mike says employers need strategic thinking, clear communication, and people who ask challenging questions.
Accountants also benefit from storytelling skills, since they need to explain information in the context of the organization’s strategy.
Blake pushes back, noting these sound like mid-level or executive skills. Mike points to internships, stretch assignments, mentorship, and scenario-based training as ways to help younger professionals practice them.
He also describes replacing an essay-based section of the Certified Management Accountant (CMA) exam with case-based questions. The goal is application. “It’s more than just knowing; it’s about doing,” he says. Still, he acknowledges some skills take time and experience to develop.
Blake suggests using AI itself for practice. A learner could describe their role and company to Claude, identify a concept they want to learn, and ask for a guided, simulated challenge.
For mid-career accountants, Mike recommends staying curious and prioritizing learning. He describes IMA’s micro-credentials, including short, focused programs in areas like AI and finance or advanced data analytics. He says professionals can learn something over the weekend and apply it at work on Monday.
Measure the results. Build the judgment.
AI’s value should show up as faster closes, fewer errors, more accurate forecasts, better decisions, and less customer friction. But those results depend on professionals who understand the business and can evaluate what the technology produces.
Listen to Episode 117 of the Earmark Podcast for Blake’s full conversation with Mike.