AI in Finance: OpenAI CFO’s Blueprint for Judgment-Driven Automation

OpenAI CFO Sarah Friar is building an AI-native finance function from the ground up, and she wants other finance leaders to understand one thing clearly: AI accelerates analysis but does not replace judgment. In a recent essay and LinkedIn post, Friar explained how OpenAI’s finance team uses ChatGPT Work to spend less time reconciling the past and more time planning the future, while keeping accountability firmly in human hands. The message resonated widely-commenters on her post emphasized that the distinction between automating analysis and outsourcing judgment is critical.

From reconciliation to real-time planning

When Friar joined OpenAI two years ago, the finance team was small and the company was scaling fast. The team set two ambitious goals: a zero-day close and continuously updated forecasting. The idea is to replace the traditional month-end scramble with a system that understands what is happening as it happens, helping leaders see choices ahead and act while the outcome can still change. Friar describes this as the real promise of an AI-native finance function: a team that operates in real time and surfaces actionable insights to the business.

Five principles for an AI-native finance team

Friar distilled her experience into five lessons for CFOs:

• Give everyone access to AI, then create a reason to use it
• Redesign workflows around the decision, not the task
• Let finance professionals become builders
• Pair speed with accountability and controls
• Measure value per unit of intelligence

The second and fifth points are the most practical for builders. Redesigning workflows around decisions means moving from automating individual tasks (like data entry) to building systems that directly support strategic choices. Measuring value per unit of intelligence shifts the conversation from cost per query to actual business outcomes.

What builders building for finance should take away

For teams building AI tools for finance departments, Friar’s approach highlights a few key design principles. First, the tool should surface insights before decisions, not just after. Second, OpenAI’s own research found that 40% of finance professionals’ specialized AI use is outside traditional finance and 22% is engineering-related-meaning tools should be flexible enough to support cross-functional work. Finally, the controls and accountability layer must be built in from the start, not bolted on after deployment. The goal is to make the finance function a strategic partner to the board, not just a faster reporting machine.

What to watch for

This blueprint comes from OpenAI’s own CFO and is based on the company’s internal experience. Other organizations may have different data residency requirements, compliance constraints, or existing system architectures that make a direct copy harder. The five lessons are vendor claims from OpenAI’s leadership, not independently verified best practices. And the Fortune article itself is a media interpretation; internal policy details may be more nuanced. For builders, the actionable takeaway is that the most successful AI finance tools won’t replace decision-makers but will make them faster, better informed, and more accountable.