OpenAI CFO’s 5 Lessons for AI-Native Finance Teams

Sarah Friar, CFO of OpenAI, recently shared five hard-won lessons from building an AI-native finance function. It’s not just about automating spreadsheets—it’s about rethinking how finance teams operate from the ground up.

For developers and founders, this is a peek into how one of the most AI-forward companies runs its own numbers. And honestly, it’s a refreshing change from the usual ‘AI will replace your job’ doom-scrolling.

Lesson 1: Automate the Forecasting Grunt Work

Friar’s team uses AI to handle the repetitive parts of forecasting—data collection, variance analysis, and scenario modeling. This frees up humans to focus on the ‘why’ behind the numbers, not just the ‘what’.

Think of it like a capybara clearing a channel through the swamp: the heavy lifting is done by the tool, so you can swim ahead and spot the caimans before they bite.

Lesson 2: AI Needs Stronger Controls, Not Weaker

With AI generating more data and insights, the risk of errors multiplies. Friar emphasizes that AI-native finance requires even tighter governance and audit trails. You can’t just trust the model—you need to verify its work.

This is a lesson for any developer building AI features: the more you automate, the more you need robust testing and monitoring. It’s like building a dam—you need to check for leaks constantly.

Lesson 3: Measure AI ROI Like a CFO

Friar suggests treating AI investments like any other capital expenditure. Track the cost of AI tools against the time saved and the quality of decisions improved. Don’t just buy AI because it’s shiny.

For a practical breakdown of AI model costs, check out our model pricing comparison to see where your budget might go.

Lesson 4: Reskill Your Team, Don’t Replace Them

The goal isn’t to fire your finance team—it’s to upskill them. Friar’s team now spends more time on strategic analysis and less on data entry. AI is the sidekick, not the hero.

This mirrors the developer world: tools like Supabase or Cloudflare Workers don’t replace engineers; they let them build faster.

Lesson 5: Start Small, Scale Fast

Friar advises starting with one high-impact use case—like automated close or real-time dashboards—and then expanding. Don’t try to boil the ocean (or the swamp) on day one.

Pick a narrow channel, prove it works, then widen it. That’s how you avoid getting stuck in the mud.

The Takeaway

AI-native finance isn’t about magic—it’s about intentional design. Whether you’re a CFO or a developer, the principles are the same: automate the boring stuff, keep control, measure ROI, and invest in your people.

If you’re building your own AI stack, you might also want to check out our reviews of Vercel for frontend, Neon for database, and Resend for email. They’re all tools that let you focus on the fun parts.

Original announcement published on OpenAI.