OpenAI used a recent webinar to demonstrate how its finance team is using artificial intelligence to transform the finance function. But while the presentation emphasized faster workflows and new ways of working, attendees repeatedly submitted questions about governance that went unanswered during the audience Q&A.
The silence was deafening.
Participants asked how ChatGPT Work addresses data governance, internal controls, agent oversight and security as finance organizations adopt artificial intelligence. Instead, OpenAI’s presenters spent the entire demo and the Q&A portion mostly discussing forecasting models, data platforms, workflow automation and the company’s new Sites feature.
The questions reflected an issue that has become one of the biggest hurdles to enterprise AI adoption in corporate finance. Before automating forecasting or accelerating the monthly close, many finance leaders first want answers to more fundamental questions: Where is the data stored? Who has access to it? How are controls enforced? Does the workflow align with SOX requirements and existing information technology general controls?
However, those concerns were not addressed by the webinar’s hosts: Kyle Kober, director of product finance at OpenAI; Stephanie Struck, OpenAI’s head of product finance; and Jackson Wang, a data scientist and member of the technical staff at OpenAI.
Guardrails before greatness
Throughout the webinar, Kober, Struck and Wang argued that AI is fundamentally changing how finance teams operate. They demonstrated how OpenAI uses ChatGPT Work to automate its own forecasting, reconcile financial models, build interactive dashboards and increase collaboration across finance.
At one point, Wang joked, “Make forecasting great again, guys,” after describing how AI had made forecasting “a lot more fun” by allowing finance teams to collaborate around live scenarios instead of being “lonely” in spreadsheets.
Yet while the presenters focused on making finance faster and more interactive, a much different conversation was unfolding in the webinar’s chat. As attendees watched the demonstrations, they asked how AI fits within finance’s existing control environment before it becomes part of the forecasting process.
Those questions, however, were never directly addressed; the Q&A discussion instead centered on topics such as OpenAI’s talent demands, data architecture and connecting to different systems, build vs. buy strategies and workflow automation.
The conversation also repeatedly returned to the quality of the data feeding AI systems. Attendees asked about creating a trusted single source of truth and using governed datasets as the foundation for AI-powered finance workflows. The questions reflected the standard CFO view that reliable outputs begin with reliable financial information.
The discussion echoed findings from a recent Intuit survey previously reported by CFO.com. That survey found that 70% of finance leaders lack a single source of truth for critical business data. More than half (57%) said delayed financial visibility caused their organization to miss a time-sensitive strategic opportunity during the previous six months.
Whether a true single source of truth is even achievable remains the subject of debate. Gartner has argued organizations should instead pursue a “sufficient version of the truth.” The questions submitted during OpenAI’s webinar show the issue remains top of mind for finance leaders as they evaluate how AI fits into areas like financial reporting and decision-making.
Limited governance references
To some extent, governance was mentioned briefly during the webinar. Kober described ChatGPT Work as “governed and auditable with controls to lock down the spend and the usage.” Later in the presentation, the speakers discussed administrator-controlled access to Sites and the use of approved “golden tables” as trusted data sources for AI-powered applications.
The webinar’s discussion mirrors what CFO.com has been hearing from finance leaders throughout the year. Previous reporting found organizations are increasing spending on AI governance, assigning internal audit teams to test AI models, investing in data architecture and more before expanding AI across finance. At the same time, broader industry surveys have found relatively few organizations believe they have the governance maturity needed to scale AI effectively.
The questions submitted throughout the webinar suggest finance leaders have already defined one of the current digital age’s most coveted questions. If AI is destined to become a part of the finance team, who is responsible for the data it uses, the costs it incurs and the systems governing it?
Bubbling AI governance concerns
The questions submitted during OpenAI’s webinar also reflect broader concerns made evident in a recent Deloitte survey. Those findings indicate that although 93% of large organizations now use AI across multiple key business functions, only 40% of CFOs said they are “very confident” in their organization’s AI governance framework.
That survey also found governance remains one of finance leaders’ biggest challenges. More than half of CFOs cited a lack of governance authority as an obstacle to building an effective enterprise-wide AI governance framework, while the most common concern was balancing pressure to deploy AI quickly with the need to manage risk.
The concerns raised in Deloitte’s research were reflected throughout OpenAI’s webinar, as presenters demonstrated how AI can automate forecasting and generate executive dashboards; attendees repeatedly steered the discussion toward governance, internal controls and data integrity.
The questions suggested that, for many finance leaders, confidence in AI’s capabilities depends as much on the guardrails surrounding the technology as the technology itself.
