It was something that thought leaders couldn’t say enough throughout Workiva’s Amplify conference: Good data is paramount for effective AI implementation.
“Bad data is AI’s kryptonite,” Alexander Davis, deputy CFO of Pie Insurance, said during a panel session. “If your organization is all-in on AI and you’re not all-in on data, you might have a problem one day. And I think a lot of folks in this room are used to sitting in meetings where data security is the topic, and I wish that collectively, we sat in rooms and worried about data quality at the same level.”
It seems many companies aren’t in a good place yet with their data. According to a recent Workiva survey, nearly two-thirds of practitioners indicated a lack of “high-quality data” for use with AI at their organizations. Practitioners who were confident in their companies’ ability to use AI were about twice as likely to have high-quality data and role-specific training compared to their less-confident counterparts.
Steve Soter, VP and industry principal at Workiva, said he was seeing those concerns about data, among other things like governance and controls, come out in conversations he was having with others at the Amplify conference.
“Yes, they’re optimistic about [AI]. Yes, they’re excited about it. But there are real challenges,” he told CFO Brew on day two of the conference.
Davis recommended companies “invest in fixing their source systems,” which can happen as they adopt AI tools—a time-consuming task on its own.
“Maybe what we do with the time we have now is get the data right, so that when it arrives, we’re ready to go,” Davis said.
Big Brother is watching you. Good data is a cornerstone of financial disclosures in a brave new AI-powered world. It’s not just the organizations using AIs to draft their disclosures; auditors and regulators, too, are using the technology to review and scrutinize them, according to Workiva CEO Julie Iskow.
Organizations should expect AI will review their disclosures even before human eyes get a first glance, Iskow said in a speech. These AI models can look for patterns, data accuracy, anomalies, and more within seconds. Importantly, the models also do not pause mid-review for feedback or follow-up questions.
“If you want to control the narrative and avoid misinterpretation, your reporting and disclosures have to be intelligence-ready,” Iskow said. Translation: “Your narrative needs to have structured data; it needs to be consistent and traceable, interpretable, machine-readable, and filled with context,” she continued.
Data alignment as an AI unlock. Adil Asar, senior director of product and engineering at Salesforce, recently helped an organization realign Salesforce’s tech stack. The result was a centralized location for Salesforce’s data, rather than its being scattered across various corners of the network. One benefit of the digital transformation project was that Salesforce’s “controllership can access and manage data seamlessly and securely across its operations,” according to a Workiva news release.
Another consequence was that the centralized data will also help “enable our agentic [AI] expansion,” Asar told Amplify attendees. “Ultimately, for any of the agents, and especially as you move in the whole autonomous universe, we wanted data which would help to remove the decision-making, and we actually think you can get the right data in so that the agents can take actions on our behalf,” he said.
This report was originally published by CFO Brew.