The AI Gold Rush Has a CFO Problem
Glenn Hopper questions AI-native ERP promises and warns about the risks of automating away the next generation of finance talent.

Sign up for smart news, insights, and analysis on the biggest financial stories of the day.
An ERP built in 18 months sounds like a venture capitalist’s dream and a CFO’s potential nightmare.
Glenn Hopper, a former finance chief who has spent the last several years helping CFOs adopt AI, took a pause when he saw a LinkedIn post from a DualEntry founder discussing how quickly the startup had built their new ERP offering.
“If I’m a CFO and I hear that, I’m nervous,” Hopper told CFO Upside on Nov. 18 at the Finance and Accounting Technology Expo in New York City. “I’m not a move-fast-and-break-things guy. I’ve got to file my quarterly reports on time, and I can’t risk a material misstatement.”
Hopper has made a career selling tools and tips for CFOs to use in the AI gold rush sweeping corporate finance. After serving as CFO at several companies, he founded RoboCFO, an advisory business helping finance teams evaluate and implement AI. He also teaches executives at Duke University, serves on advisory boards focused on technology and accounting and has written books on the subject, including his newest, “The AI-Ready CFO”.
So what concerns someone who has spent years championing AI about the software CFOs are buying and the next generation of finance leaders?
Faster Software, Slower Trust
The ERP arms race was on full display at FATE, where Campfire, Rillet, DualEntry, Lensing and ianaiERP exhibited alongside established providers Oracle NetSuite, SAP, Sage and Intuit Enterprise Suite. Hopper likens the challenge facing incumbents to rebuilding a battleship while it’s underway, with established vendors responsible for maintaining systems companies depend on for financial reporting.
“You can’t break your product for all these public companies that are on your product and are relying on your system,” he said.
Startups can design technology around AI agents without maintaining decades of existing infrastructure, giving them an advantage in introducing new capabilities and shortening implementation timelines, Hopper explained. “The functionality in them is more natively modern, and there’s a lot of advantages to it, and especially around the quicker installs that can be done,” he said.
Still, he suggested startups might build trust by initially offering narrower products, such as accounts payable or accounts receivable automation, before attempting to replace an entire ERP. “You could earn trust slowly, but the hurdle is so high that you have to convince a CFO that you’re ready to swap out something that he’s been familiar with for 30 years,” he said.
For the CFOs evaluating software right now, Hopper believes the most important work happens before meeting vendors, when finance teams identify data silos and determine which problems need solving. “You can’t buy the product first and then figure out how to jam a square peg into a round hole,” he said. “You have to figure out what you are solving for, and then you go ask the vendor, can you do this, and how would you do this?”
Hopper sees yet another challenge as AI automates the work traditionally assigned to junior accountants. It entails companies eliminating the work that develops future finance leaders. “Nobody goes and gets a master’s in accounting or a master’s in finance because they love data entry and they love swivel-chairing data,” he said. “But it was sort of a necessary evil.”
As automation takes over those entry-level responsibilities, Hopper believes companies must deliberately replace the learning experience. “We can’t just mindlessly roll through and watch automation and AI sort of eat up those bottom layers because then we’re going to cripple ourselves at some point once the top people age out and then we don’t have anybody to replace them,” he said.
Hopper, an advisory board member at the University of Memphis’ Crews School of Accountancy, envisions apprenticeship programs that rotate junior employees through sales, operations and other departments. New hires could also review AI-generated work, with supervisors deliberately introducing mistakes to develop their judgment. “You could even, in a training program, put in incorrect things or whatever, and see if they catch them,” he said. “And if they don’t catch them, it’s like here’s a training moment.”
Such programs require investment when executives increasingly expect automation to reduce costs and produce measurable returns. “Everybody wants to show ROI on AI investments, and if you’re bringing people in and you don’t have work for them to do, it’s like, well, where’s our return on investment coming from the AI?” Hopper said.
He expects accounting knowledge to remain essential for future CFOs, with systems thinking, data science and strategy becoming increasingly valuable. “The differentiator is going to be, yeah, no one’s excited about someone who just closes the books right every month,” he said. “What insights can you give me?”
The Control Freaks’ Head Start
One of Hopper’s biggest surprises about AI adoption involves the financial controls he initially expected to complicate implementation.
GenAI’s probabilistic nature raised questions for accounting teams accustomed to producing consistent, reproducible results. “I need the trial balance to balance. I need the same numbers every time,” Hopper said. “If AI generates a number in March, I need when the auditors come and look at it for our annual audit, it needs to generate the same number.”
Agentic AI has changed his thinking because companies can combine AI models with things like conventional code, predefined workflows and detailed logs. Finance departments already maintain procedures governing reconciliations and monthly closes, providing a framework for developing agents. “The surprising part of that is I thought that these workflows were going to slow us down,” he said. “But now with agents and agent harnesses, you can make it do the same thing every time and log everything it’s doing.”
Hopper, who helped reduce the monthly close from 30 days to 10 days through back-office automation while serving as COO and CFO at CICAYDA, believes established financial processes offer an advantage when designing reliable AI systems. “When I come in and talk to finance teams, they already have the playbook that we’re going to use with the agents,” he said.
Hopper also worries about employees using sophisticated analytical tools without understanding their conclusions. Asked about cybersecurity, he acknowledged that additional software creates vulnerabilities, although he considers employees misinterpreting AI-generated analysis a particularly important risk. “The bigger fear is people who don’t understand the data that they’re working with, drawing weird conclusions that they can’t back up, and using AI to support their conclusions,” he said.
He recalled working with DataRobot, a machine learning platform that allowed employees to generate analytical models through a drag-and-drop interface. Some users accepted results without understanding the underlying algorithms.
“You’d ask them, ‘What algorithms did you use? How did you come up with that?’ And they don’t know, but they’re treating it as gospel because this tool put it out,” Hopper said.
Technology purchasing decisions create another opportunity for things to go wrong, particularly when CFOs advance implementations without involving their CIO until late in the process. “If they don’t loop the CIO in early, they’ll come in in the 11th hour and be like, ‘You’re talking about doing something with systems and with our data and everything, and you haven’t even brought me in here,’ and they’ll kibosh it,” Hopper said.
Go Touch Grass: Hopper also sees value in finance leaders comparing their experiences with peers as vendors showcase increasingly ambitious AI capabilities. At conferences like FATE, he said, CFOs want to understand whether their companies are moving too quickly or falling behind. Those conversations can provide a useful reality check when the technology industry’s promises are moving faster than companies’ ability to evaluate their results.
“What should I be scared of? What should I be doing with it? What are you doing with it?” Hopper said, describing the questions finance chiefs bring to those conversations. “If you’re just arguing with your chatbot all day and trying to figure stuff out, it can be this really weird, bleak existence.”











