Problem solved: A math example of CPA-led assurance using AI agents

At first glance, Richie Sater, CPA, had no business solving a problem that had stumped mathematicians for four decades.

But Sater actually may be the perfect person, and accounting may be the perfect profession for solving a different problem that has stumped mathematicians for even longer.

Sater, a technology entrepreneur, recently joined the AI-driven craze of solving seemingly unsolvable math problems by cracking the code on a query that a Belarusian mathematician entered into a repository for unsolved problems in 1986.

After an army of AI agents that Sater built came up with an elusive answer over the course of about three weeks, Sater unearthed another problem that needs solving in the math community — in an area of expertise for CPAs.

“They have a huge reviewer bottleneck and need more scalable ways to verify AI-generated work,” he said.

Sater believes the solution to that problem is the CPA, and he sees an opportunity for enterprising CPAs to apply their existing assurance skills to AI-generated work in mathematics, financial analysis, and other business processes.

Opening doors with new tech, old skills

“I’m a CPA and a software developer — a bit of a weird Venn diagram to be in the middle of,” said Sater, 32, who was introduced to coding at an early age by his mother, a programmer at the Pentagon. “I asked my parents, ‘What’s the most math-oriented business major?’ And they said, ‘Accounting.’ I learned later it’s economics or maybe finance, but I majored in accounting and I got my CPA, and it ended up being great.

“The CPA has been so wonderful for me. It has opened so many doors.”

Now Sater is looking to open a door for himself and other CPAs from what he’s gleaned while “puttering around” the world of advanced mathematics. After his AI agents solved Problem 10.34 in the Kourovka Notebook (“Finite groups that are the product of every pair of non-conjugate maximal subgroups are soluble”), Sater contacted mathematician Viktor Monakhov not merely with an answer but also with supporting materials.

“I solved it using AI, but I didn’t stop there. I put an audit package together for it, and I showed the audit package to the mathematicians,” Sater said. “So instead of having to spend two weeks like they do with every other person who says they have an answer, they could just quickly look at the audit package. And it got endorsed and sent to peer review.

“Monakhov got back to me and said, ‘This is great,’ and he gave me another problem. I solved that. He’s giving me more problems. At this point, I’ve got several papers submitted for publication.”

The last two, Sater said, took about three days instead of three weeks for his AI agents to solve.

Sater is now a part of what he called a “gold rush” to solve long-unsolved math problems, powered by AI. That has created another problem for the math community.

“They are generating all of this work with AI, far more work than they used to,” Sater said. “They have a peer review process, but it’s not equipped for assurance at a glance. It takes a year to get something through peer review.”

That’s where CPAs can step in. Sater estimated that a CPA with basic computer literacy can start building basic AI agents in about a week. For Sater, mathematics is a concrete demonstration of a broader opportunity: helping organizations evaluate AI-generated analyses, financial models, and other business outputs.

“I was really shocked at how welcoming mathematicians are to CPAs,” Sater said. “My internship was in internal audit, and nobody’s excited when the internal auditor hits their cubicle. This has been the complete opposite.

“There’s an assurance crisis happening in math right now, but the same challenge applies to businesses using AI. That creates a business opportunity for CPAs to help companies verify AI-generated work before they rely on it.

“The door is wide open for this, and we’re already the people who provide assurance.”

— To comment on this article or to suggest an idea for another article, contact Bryan Strickland at Bryan.Strickland@aicpa-cima.com.