Sarah Friar has a clear message for anyone eyeing a finance role at OpenAI. Mastery of tools such as Codex now stands alongside spreadsheet proficiency as a baseline requirement. “I would never hire a finance person who didn’t know how to use Excel, and I probably wouldn’t hire a finance person today that doesn’t know how to use a tool like Codex,” she said in a recent interview.
The remark landed with force. Friar, who joined the AI company as its first chief financial officer in 2024 after leading Nextdoor and serving as Square’s finance chief, spoke at the Liquidity Summit. Her words reflect a sharp turn in expectations for financial professionals. They also signal how deeply OpenAI itself has embraced its own technology to reshape internal operations.
Codex began as an AI coding agent. It has evolved. More than five million people now use it weekly. Knowledge workers — analysts, marketers, operators, investors, bankers — account for about 20 percent of those users. That group grows more than three times as fast as developers. OpenAI’s announcement on June 2 detailed the shift.
Inside OpenAI, non-technical teams deploy Codex to build internal applications, prepare executive materials, generate dashboards and transform creative briefs into brand-aligned output. The pattern repeats across functions. Finance teams at the company and beyond stand to gain the most from the latest expansions.
On Tuesday OpenAI rolled out a series of plugins that adapt Codex to specific roles. The public equity investing plugin lets investors review earnings, compare companies, track market signals and test investment theses. It pulls from Moody’s, Daloopa, Datasite, FactSet, LSEG, S&P, PitchBook and Hebbia. The investment banking plugin helps bankers assemble pitch materials, analyze comparables and convert diligence findings into client recommendations.
Additional plugins target data analytics, sales, creative production and product design. They connect directly to Snowflake, Databricks, Salesforce, HubSpot, Figma, Canva and more than 60 other applications in total. Corporate finance, legal, private equity and strategy consulting plugins will follow soon. OpenAI also plans to fold Codex capabilities into its flagship ChatGPT product. Bloomberg reported the competitive push on June 2.
The timing carries weight. Anthropic has released its own set of AI agents aimed at banks and financial services firms. Those tools handle tasks from drafting pitch decks to reviewing financial statements and managing compliance. The race for enterprise adoption intensifies as both companies eye eventual public listings.
Friar’s finance organization has expanded in parallel. She hired Ajmere Dale as chief accounting officer and Cynthia Gaylor to lead corporate finance, long-range planning, capital strategy and investor relations. The moves address the demands of scaling a business that burns through enormous sums on computing infrastructure while preparing financial statements suitable for public markets. CFO.com covered the appointments in January.
Preparation for an IPO consumes much of her attention. The Wall Street Journal detailed last month how Friar has hired a wave of new finance and accounting professionals and begun informal talks with banks. She has moderated some of the company’s more ambitious spending projections. Sam Altman once referenced $1.4 trillion in future compute obligations in a public presentation. Friar later clarified the figure closer to $600 billion through 2030. The Wall Street Journal reported on May 2.
Yet compute scarcity remains the binding constraint. “Compute is a very scarce resource at the moment,” Friar said. Demand climbs a vertical wall that continues to outpace supply. Even with heavy upfront investment, OpenAI expects the shortage to persist through the rest of this year. “Thank God we did,” she added, referring to earlier infrastructure bets, “because in ’26 we still won’t have enough compute.”
The numbers tell part of the story. OpenAI’s computing capacity expanded from 0.2 gigawatts in 2023 to 0.6 gigawatts in 2024 and roughly 1.9 gigawatts last year. Plans call for locking in around 30 gigawatts of power over the next five to seven years. That scale matches dozens of large nuclear plants. Revenue projections tied to each gigawatt range from $20 billion to $40 billion annually at maturity. The infrastructure decisions now shape the company’s cost structure for years ahead.
Finance leaders elsewhere have taken notice. Deloitte’s Finance Trends 2026 survey of more than 1,300 global finance executives ranked AI and automation skills as the top priority for talent development and sourcing. The category outranked traditional strengths such as regulatory compliance and cost management. The message aligns with Friar’s stance. Those who fail to adopt these tools risk falling behind.
But adoption brings its own complications. Many organizations still operate with a mismatch between AI capabilities and actual value captured. Advanced models sit lightly integrated into workflows. Real productivity gains require redesigning processes around agentic systems that can act across applications, interpret context and iterate without constant human prompting.
Friar has witnessed the internal variation firsthand. When she arrived at OpenAI she asked her finance team how they used AI tools. Responses ranged from enthusiastic experimentation by a few to near-total absence among recent hires from traditional firms. The gap prompted her to push for structured adoption. The results appear in faster contract processing, automated investor interactions and self-improving agents in areas such as tax preparation.
One collaboration with Thrive Holdings and Crete, a network of accounting firms, produced a tax agent built on Codex. The system automates filings, learns from production outcomes and improves accuracy over time without engineers chasing every error. Similar loops could transform financial close processes, audit preparation and forecasting.
Yet the technology does not replace judgment. Friar still emphasizes trust, community relations and sound capital allocation. Her background informs that balance. Raised in Northern Ireland during the Troubles, she earned an engineering degree before moving into strategy consulting at McKinsey, technology equity research at Goldman Sachs, the Square IPO and the Nextdoor SPAC. Those experiences taught her how to guide high-growth companies through complexity.
She now applies the same discipline to OpenAI. The company must balance explosive product demand against enormous capital needs. It must satisfy investors while preserving its original research mission. And its finance team must model the very AI fluency it demands from new hires.
The bar has moved. Excel fluency once defined competence in finance. Today that standard includes the ability to direct AI agents that write code, pull data across systems, generate reports and recommend actions. Tomorrow it may demand fluency with autonomous agents that handle entire workflows from forecast updates to board presentations.
Friar’s comments serve as both warning and invitation. Finance professionals who ignore the shift will find fewer opportunities at leading AI organizations. Those who embrace it may discover their roles expand from scorekeepers to strategic partners capable of shaping business outcomes at unprecedented speed. The companies that integrate these tools most effectively will set the competitive pace for the next decade.
And the pressure only grows. With Codex plugins now targeting corporate finance and legal functions directly, the next wave of automation stands ready to enter the office of the CFO. How finance leaders respond will determine whether they ride that wave or watch it pass them by.
OpenAI’s CFO Sets a New Bar: Finance Talent Must Master AI Agents Like Codex first appeared on Web and IT News.