Companies are using AI wrong, EY’s Deirdre Ryan says

Many companies aren’t “seeing the game-changing effect[s]” of agentic AI that they expected, because they’re “not optimizing” how they use the technology, according to Deirdre Ryan, global finance transformation leader at professional services firm EY.

Companies are “apply[ing] agentic AI to existing processes” and “tweak[ing]” them, according to Ryan, who estimates she speaks with two to three CFOs a week. While this may drive some “operational efficiency,” Ryan recommends that companies reimagine what they’re trying to accomplish. Then, determine the tasks needed to achieve this result, and “translate these jobs to an agentic finance workforce” with humans in the loop, Ryan told CFO Brew.

“What we’re saying to clients is that you don’t want to do things differently. You want to do different things—and what do we mean by that? What we mean is, understand the capabilities of these disruptive technologies and build in AI,” Ryan said.

Less strategic. What’s true for agentic AI is true for AI investment generally—the focus tends to be process efficiency. In a Gartner survey of more than 200 finance leaders conducted in March, 45% of respondents said their AI investments in finance lean toward productivity, while 20% said theirs lean toward decision quality.

“Many CFOs are prioritizing AI use cases focused on productivity and efficiency. However, boards place greater emphasis on investments that drive growth, improve decision-making, and deliver competitive advantage,” Shankar Keshav, principal analyst with Gartner Finance, said in a statement commenting on the survey in July.

AI use cases tend to be concentrated in areas “that enhance individual productivity or streamline transactional processes,” but “there is a ceiling to the benefits they return for most organizations,” according to Gartner.

RPA redux? The current use of AI is similar to the advent of Robotic Process Automation, or RPA, Ryan said. Companies typically use RPA to automate repetitive tasks like data entry or invoice processing, according to finance and HR platform company Workday. RPA has been around since the 1990s, but mainstream adoption didn’t occur until the 2010s.

The RPA rollout was haphazard. Companies would “put a little RPA here and a little RPA there,” Ryan said. Sometimes, this would produce “a bit of efficiency; maybe they saved an FTE [full-time equivalent]. It really was death by a thousand cuts, and they didn’t see that overall impact.”

Ryan, who runs EY’s global and US finance consulting practice, has spent years in professional services. She joined EY in 2023 after working for 25 years at Deloitte. Ryan said she realizes that the transition to AI can be difficult for high-level executives. “It’s exceptionally hard for very tenured people who have been doing a process a certain way for a really long period of time” to change, she said.

No standouts. When asked which companies have succeeded with AI, Ryan said she couldn’t think of one. She’s not referring to AI-native companies but to large, traditional firms—those with global footprints and “multiple underlying ERPs…I don’t think there is a company out there that has completely transformed every aspect of their finance organization with AI yet,” Ryan said.

“There are a significant number of companies” that have identified areas in finance where they’ll see “significant impact or value” using agentic AI, she said. This includes areas of finance like record-to-report and FP&A.

One area where agentic AI can transform a process is scenario planning, which organizations have historically used to evaluate the impact of unforeseen events, Ryan said.

When global companies wanted to assess the possible financial “impact of a specific scenario,” some firms would send queries to their senior finance leaders, and wait for them to respond in Excel, Ryan said. Then, the company would have to analyze and aggregate the information, while making sure to screen out errors. The process could take weeks. “It’s such a manual, laborious process.”

But companies with AI could play out scenarios “effectively instantaneously,” according to Ryan, noting that firms would need to have the correct data to do it.

“Agentic AI has the capability to drive an incredibly accurate predictive forecast for an organization based on drivers,” Ryan concluded.

—Additional reporting by Demi Lawrence