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

Praveer Melwani’s central argument is that finance should act as a business leader, using data and automation to shape decisions rather than serving as a “traffic cop.” He joined Figma in 2017 as its first business operations hire and later became CFO. The role, he says, requires understanding the business, applying operational rigor, driving change, and aligning people—work that depends more on communication, relationships, and influence than he initially expected.

At Figma, Melwani used SQL to connect product and utilization metrics with Stripe financial data, then worked with data scientists to add those tables to recurring ETL. Consistent definitions helped teams move past arguments about what was true and focus on performance; he estimates that getting data into the right place accounts for 95% of the challenge. He reviews hourly dashboards or daily metrics with business counterparts to identify anomalies and decide whether action is warranted. A pandemic-era usage spike took weeks to diagnose, he recalls, whereas similar questions can now be answered in about an hour. Early detection can save hundreds or thousands of dollars when a trend is unfavorable, or support additional investment when it is favorable.

AI is helping Figma’s finance teams automate repetitive work and spend more time on decisions. Credit and collections team member Kajal rebuilt invoice tracking, collections, and reconciliation workflows with NetSuite, freeing time to work on reducing days receivable and automate further tasks; she also made reusable tools for credit checks and combining datasets. Corporate development uses AI-generated product recreations to assess a target’s technology and how difficult its product might be to reproduce, in partnership with Figma’s product team. AI personas of Melwani, Dylan, and CRO Sean review investment memos to surface likely questions and diligence gaps, reducing the time leaders spend reviewing documents together. For travel-and-expense planning, corporate finance combined Navan, SIP, and Ramp data. Fuzzy matching and data preparation were substantial hurdles, but the suggestions were useful and directionally correct, with feedback visible in visualizations 30 minutes later.

Melwani sees finance tools moving beyond tracking workflows toward delivering results while preserving an audit trail. AI can provide initial drafts, expectations, and investigative leads; tools that flag general-ledger changes can give accountants a starting point for flux analysis, even if they do not immediately identify the error. Deterministic processes could be fully automated when they have enough information to produce trustworthy, audit-ready outputs, but business-critical automation depends on auditors’ confidence. He says current AI models are not good enough to create forecasts or financial models, though they can help identify a complex model’s key drivers and sensitivities. Excel remains valuable because directly exploring a model can build intuition and trust; prompting is useful, but does not replace testing ideas in the model.

Figma is in what Melwani calls a “messy middle”: some teams have centralized support for AI-built tools, while others have moved faster than the infrastructure. He says CIO teams are needed to manage dependencies and hosting and provide reusable foundations. Figma’s new CIO, Casper Sorenson, joined from Coinbase, and Melwani discussed centralized project-management support to help business teams scale successful tools. He favors buying core systems and systems of record when reliability is critical, while allowing experimentation with AI for workflows. That experimentation has made work more enjoyable, and team members have said it makes them want to stay. He wants AI to give finance staff more time to find and strengthen valuable opportunities, not replace their judgment; he also warns that AI-generated documents can burden readers and that low-quality output is increasingly called “slop.”

Melwani looks for intellectually curious hires who understand the business, reason from data, and communicate clearly. In a VP Finance case study, he assesses candidates’ use of tools, grasp of model sensitivities, business intuition, and communication. He expects finance to become more efficient, but says new products, acquisitions, and business evolution keep adding complexity; making operations more consistent is necessary to sustain efficiency. He has suggested that if complexity stayed constant, AI could eventually mean finance needs fewer people, while also describing a near-term possibility of maintaining the same headcount as efficiency improves. Reliable data is a prerequisite: Yvonne advises companies to establish data integrity, access, and tracking before prioritizing close processes or board updates. Melwani says investor and board analysis is becoming deeper and faster, and companies are expected to explain their businesses and numbers earlier, including in Series A and Series B fundraising.