The New CFO: How AI Is Rewriting the Finance Function | Figma CFO Praveer Melwani

Figma's CFO rebuilt collections and reconciliation around AI

Figma's finance function is being rebuilt around a bet that AI should change what finance people spend their time on, with speed as a byproduct. Collections and reconciliation are automated, corporate development prototypes acquisition targets and tests memos against simulated versions of Figma's own leaders, and the remaining people act as business partners. CFO Praveer Melwani says the hiring bar has not moved.

Takeaways

• AI shifts what finance people spend time on. Melwani wrote 400 to 500 line SQL queries in his early Figma days, joining product metrics with Stripe data. He says AI should change what finance people work on, moving the function from traffic cop to business enabler.
• Figma automated collections and reconciliation. Figma's credit and collections lead rebuilt invoice tracking and cash collections into a homegrown system that drives down days receivable. Corporate finance reconciled Navan, SAP and Ramp data and delivered a visualized analysis 30 minutes after Melwani's feedback.
• Corp dev uses AI prototypes as diligence. The team asks whether a target's product can be vibe coded, and runs investment memos past AI personalities modelled on Figma leaders including Melwani and CRO Sean, cutting the time leadership spends reviewing together.
• The hiring bar is unchanged. Melwani is hiring a VP of finance and keeps the case study sacred, testing whether a candidate understands the model they built and can explain it. He wants intellectual curiosity and data fluency, and says you cannot outsource your thinking to AI.
• Bottom-up builds need productizing to last. A growth-stage CFO told Seema Amble his team built 40 or 50 agents in a month or two, and about 90% went unused. Figma hired Casper Sorensen from Coinbase as CIO to harmonise and productize bottom-up builds so they outlive their creator.
• Reporting cadence moved from monthly toward daily. Makarov says cloud data moved reporting from monthly or quarterly to weekly, and daily is now possible. A COVID-era usage spike at Webflow took weeks to diagnose and would take an hour today. Storytelling now matters as early as Series A.

What comes next

• Melwani expects deterministic processes to become fully automated with audit-ready outputs. He will add finance headcount only when complexity grows faster than the need to support it, and says flat complexity probably means Figma would not need the same number of people a year from now.
• Amble sees the finance engineer spreading across teams, a data scientist with accounting grounding who owns metric definitions. Melwani insists definitions stay consistent and historicals unchanged.