Lessons I learned from building a finance function around AI

If you work in finance and you're wondering how AI will actually reshape your day-to-day work—not in some vague future, but right now—the experience of OpenAI's own finance department offers a useful case study. OpenAI's Chief Financial Officer, Sarah Friar, has shared five practical lessons about building what she calls an "AI-native finance function." Her insights cover everything from automated forecasting to tighter internal controls and figuring out the real return on AI investments. Here's what finance leaders and team members can take away from her approach.

Why Finance Teams Can't Afford to Wait

Finance has always been a data-heavy discipline, which makes it a natural fit for AI tools. Yet many departments are still stuck in manual mode: copying numbers between spreadsheets, chasing approvals over email, and spending days assembling reports that AI could draft in minutes.

Friar's core argument is that finance shouldn't just adopt AI tools one at a time. Instead, teams should rethink how the whole function operates when AI is baked in from the start. That means designing processes, controls, and team skills around the assumption that AI handles a meaningful share of the analytical and administrative workload.

For readers, the takeaway is a practical roadmap: five lessons drawn from a finance organization that is living through this transition at one of the fastest-moving AI companies in the world.

Lesson 1: Automate Forecasting, But Keep Humans in the Loop

The first lesson centers on forecasting. Traditional financial forecasting is slow and often lags behind reality. AI changes the math by continuously ingesting data—sales figures, spending patterns, market signals—and producing updated projections far more frequently than a monthly or quarterly cycle allows.

Friar's point isn't that AI should replace the judgment of finance professionals. Rather, automation should handle the heavy lifting of data processing, while people focus on interpretation, assumptions, and decisions. The human role shifts from number-cruncher to reviewer and strategist. Teams that get this balance right produce forecasts that are both faster and more thoughtful.

Lesson 2: Strengthen Controls, Don't Loosen Them

A common worry as AI enters financial workflows is that oversight will slip. If a model drafts a report or flags an anomaly, who verifies it? Friar's answer is that AI-native finance actually allows for stronger controls, not weaker ones—provided you design them deliberately.

When systems are built to log AI-assisted actions, flag unusual transactions, and escalate edge cases to humans, the control environment can become tighter than it ever was under manual review. The lesson for finance leaders: treat AI governance as a core design requirement, not an afterthought. Build in checkpoints, audit trails, and clear ownership of AI outputs from day one.

Lesson 3: Prove the ROI of AI—On AI Itself

It would be ironic for a finance team to adopt AI without measuring its value, and Friar doesn't let that happen. One of her lessons is to apply the same financial discipline to AI investments that finance teams apply to any other spending.

That means tracking what AI tools actually save or generate: hours reclaimed from manual tasks, faster close cycles, fewer errors, better decisions. Not every AI investment will pay off, and honest measurement helps teams double down on what works and quietly retire what doesn't. For finance departments, this is familiar territory—apply your own toolkit to the technology you're adopting.

Lesson 4: Rethink Team Roles and Skills

An AI-native finance function changes what people do all day. Routine reconciliation and report assembly shrink, while work that demands judgment—scenario planning, business partnering, communicating insights to leadership—grows.

Friar's guidance suggests finance professionals should lean into that shift. Building comfort with AI tools, learning to write clear instructions for them, and developing the critical eye to review AI outputs are becoming baseline skills. Leaders, meanwhile, need to invest in training and reassure teams that the goal is to elevate their work, not eliminate their roles.

Lesson 5: Move with Intention, Not Haste

The final lesson is about pace. The temptation is either to jump on every new tool or to wait until everything is settled. Friar's example points to a middle path: start with high-value, lower-risk use cases, learn from what works, and expand deliberately.

This phased approach lets teams build confidence, refine controls, and demonstrate early wins that build support for bigger changes. Waiting for perfect clarity means falling behind, but reckless adoption creates risk. The AI-native path is iterative.

Putting the Lessons to Work

For finance teams starting out, a sensible first move is picking one workflow—monthly reporting, expense review, or cash forecasting—and piloting AI assistance there, with clear metrics and human review built in. From there, expand based on results.

Friar's overall message is optimistic but grounded: AI won't replace finance professionals, but finance functions that restructure themselves around AI will simply operate at a different level of speed and insight than those that don't. The five lessons—automate forecasting with human judgment intact, harden controls, measure AI's ROI honestly, reskill teams, and move with intention—offer a clear starting point for any finance organization ready to make that leap.