Both ChatGPT owner OpenAI and Claude operator Anthropic have disclosed cyber incidents involving their AI tools in recent months, CFO Dive reported, prompting finance chiefs to take second looks at the controls they have in place. The last thing CFOs or chief accounting officers “want to see is a rogue agent” inside of their record to report or financial close process, Trintech CFO Omar Choucair said.
With the possibility of a rogue AI agent floating around, “the question then is: Does the company have the controls and the review processes in place?” Choucair told CFO Dive in an interview. “I think because a lot of companies don't have that control defined and enacted and implemented, that's why they've slowed down on the traction on the agents.”
Getting off the AI starting block
“Rogue” AI agents, or agents which perform assigned duties in unexpected ways, have captured a widening amount of business leaders’ attention in recent months as debates on AI regulation and security sharpen.
That’s added another layer of complexity to the AI question for CFOs. Many finance chiefs are still mulling where and in which processes AI agents could be most beneficial inside of key financial processes such as the record to report cycle — which aggregates, processes, and then reports financial information from various sources.
As a technology provider — offering AI-enabled software to help finance teams automate the financial close — Trintech has seen the full gamut of AI integration and readiness as it evolves its own AI tools and agents, Choucair said.
There are some companies which are “charting new territories, and there are companies that are still in the starting blocks,” Choucair, who has served as CFO for the Plano, Texas-based company since 2019, said.
For CFOs that have yet to hop on the AI fast train, that lack of clarity into exactly what an AI agent is doing is likely a key reason, he said.
“I always say that CFOs get hired because they're good at what they've done: They have a track record of what they've done [and] they get fired…if they miss a number,” Choucair said. “So I think there's just a lot of resistance and hesitancy, because people know what they get hired for, and they know what they get fired for.”
Ungoverned AI tools have also led to rising costs as well as security or reporting challenges. A recent survey by Redwood Software found 77% of senior tech, risk and AI leaders said their organizations had experienced negative consequences from ungoverned AI agents, citing unexpected cloud costs, operational disruption, and compliance issues, CFO Dive previously reported.
Finance chiefs learned that lesson quickly, however: Many companies may not have had the necessary controls needed to track token usage, for example, but they adapted quickly after getting stung by high costs for token overuse, Choucair said.
“Fool me once, but don't fool me twice,” Choucair said of CFO’s current perspective on AI costs and ROI. “Once you get this huge bill, it's not going to happen again. Most CFOs are going to say, ‘Okay, well, I know how I'm going to stop that.’”
Right team, right tools
Another area that could be causing friction is a lack of compatibility between the AI tools and finance team members, he said.
“I think a lot of executives are looking around, going ‘Oh my gosh…I don't know if I have the right team to go through a financial transformation with new technology,’” Choucair said.
Many companies are “frustrated because they’re trying to bring in all the technology and the new digital transformation platforms, et cetera, and they're realizing that they're struggling: Their people are struggling, because they may not have the right skill set,” he said.
In the next 12 to 24 months, Choucair sees “a new wave of people coming in that are [data] savvy and that will enjoy that work as opposed to running from it.”
That wave will occur as AI itself continues to evolve, moving from tools that require significant oversight to technology that can run almost independently.
Trintech recently launched three new AI agents — a data access, accruals intelligence, and exception management agent, respectively — focused on the financial close process, according to a Sept. 28 press release. The three agents, which join two additional AI agents Trintech launched earlier this year, mark a shift from “AI that surfaces insights to AI that executes the work itself,” the company said in the release.
Choucair stressed that these more autonomous technologies still require the proper guardrails.
“The whole idea is that time savings [from the agent] is spent reviewing those documents and just getting comfort that what [the agent] did…makes sense and it is reviewed,” he said. “The better the data, the better the controls are. Then yes, those agents will have better agency. It’s just not quite there yet.”
