A solid data foundation is the CFO’s business

AI is a powerful tool, ready to deliver CFOs all sorts of insights on a company’s financial performance, nearly instantly. The problem for many, though, is the quality of the data (or lack thereof) those AI tools are drawing from. Indeed, only 11% of executives believed their data quality was sufficient for AI, and more than a quarter (27%) said poor data quality was blocking AI deployment in certain workflows, according to a recent Workiva survey of nearly 2,300 finance, risk, and sustainability pros globally.

A solid data foundation—one that has “complete, accurate, reliable” data—allows organizations “to really maximize the use of technology tools and also AI,” Martino Cadoni, CFO of language translation platform DeepL, told CFO Brew. An organization “can have the best AI tools in the world,” but that doesn’t mean much if their output is based on bad data.

“As a CFO, you really need to ensure that before you implement all these very sophisticated and fancy AI tools, you have a reliable, strong data foundation, which then enables you to trust whatever AI does,” Cadoni said. “If you don’t have confidence in the data layer, then everything else is really questionable.”

Building blocks. Companies need to set appropriate internal controls to check the accuracy and completeness of their data, Cadoni said.

DeepL has a data lake that holds vast quantities of its information. This centralized location allows DeepL to consolidate its disparate tech systems. But within this data lake, the company has a “semantic layer,” he said—“a set of data and definitions that is approved and signed off [on] by relevant owners to ensure completeness and accuracy.

Cadoni works with “the relevant teams to build a regular, frequent data flow from our source systems into the data lake.” Finance’s internal controls team and the operations function’s data team work together on “IT controls, reconciliations between source systems and the data lake, definition reviews, and monthly management reviews to ensure the data makes sense and that we understand the trends.”

Cadoni also sits in on quarterly meetings with the company’s data-lake vendor “to make sure that we are aligned on data strategy and are happy with the services provided.”

Clean freaks. CFOs are “hypersensitive” to the need for clean data because they rely heavily on it to read and interpret company finances, according to Kevin Carmody, senior partner at McKinsey.

And why not? CFOs are the “voice of the company” in communicating financial information to all stakeholders, from investors to board members, Carmody told CFO Brew. Beyond that, finance leaders rely on data to determine “where the business is headed” to help set strategy.

Degrees of freedom. In the early days of generative AI (which wasn’t that long ago, Carmody reminded us), “the constraint was the data; they didn’t have clean data, so they couldn’t actually get at that in a way where they were comfortable that they could analyze it in an accurate way that provided meaningful insights to the leaders across the business.”

“And I think that has changed,” Carmody said. Now, CFOs can work with AI agents “almost as an employee,” which he called a “game-changing element.”

“What that’s basically done is…created degrees of freedom for CFOs to look at their business differently and move with speed,” he said. “The insights they’re able to glean if they effectively use agents, how they upskill their workforce to get the better answers faster, how they compete against their peer group, where they want to stay one or two steps ahead—I think that is where the exciting part of AI intersects with the CFO’s job on a daily basis.”

With powerful AI tools at his disposal—and confidence in their output, thanks to DeepL’s data foundation—Cadoni said he can call up detailed financial information in a fraction of the time it took when he started his career. Back then, you’d have to call up the FP&A manager if you wanted revenue numbers for a specific country, who’d then perhaps have to get another person involved to query that data.

“I get the data faster [and] it’s less distractions on my teams, so now they can be focused on other things rather than being reactive on my request,” he said. “So it’s a win-win.”