How OpenAI's Finance Team Does 100% of the Work with 20% of the Headcount

Introduction: What Does an AI-Native Finance Team Look Like?

In a recent sharing session at OpenAI, Stacie Faggioli, Head of Applied Business Finance, gave a detailed look at how her team has used AI tools to fundamentally reshape financial workflows. One striking data point stands out: according to a PwC assessment, OpenAI's finance team is only 20% the size of comparable tech companies — yet delivers the same or even greater output.

This isn't a story about "replacing people with AI." It's a case study in how to embrace AI holistically — from organizational structure and work philosophy to specific tools.

Three Core Principles: Building an AI-Native Finance Team

AI Native by Design

Stacie emphasized that when the team deploys AI tools or Agents, they never treat them as "add-ons" to existing processes. Instead, they fundamentally reimagine how workflows should operate. Specific practices include:

• Spreading best practices across the entire team and cross-functional stakeholders
• Rethinking hiring strategies and organizational design around AI Agents
• Embedding engineers directly into the finance team (rather than relying on IT), so tools and business operations evolve together in real time

Demonstrating Headcount Leverage

PwC's assessment revealed that OpenAI's finance team is roughly 20% the size of comparable tech companies. This is living proof that "with the right technology and tools, you really can do more with less."

Deploy Early, Iterate Fast

When facing the decision of "should we wait for a more stable version," the team's bias is always to deploy first and iterate later. This allows the team to grow organically alongside the technology, rather than passively waiting for the perfect solution.

Organizational Innovation: Engineers Embedded in Finance

OpenAI's finance organization is built on three pillars:

1. Strategic Finance: Responsible for capital allocation, fundraising, and growth planning
2. Finance Operations: Covers accounting, revenue collection, tax filing, monthly close, and other foundational work
3. Enterprise FinTech: Owns financial systems and data platforms

The key innovation lies in the third pillar — embedding engineers who build AI tools and Agents directly within the finance team. Stacie explained that having engineers work side by side with finance domain experts dramatically accelerates deployment and iteration, rather than "sitting in a corner waiting for IT to build to spec."

Personal Productivity Tools: Three Real-World Use Cases

Investor Relations Agent: Saving Hundreds of Millions in Advisory Fees

Over the past year, OpenAI completed two historic fundraises — a $40 billion private funding round and a $122 billion raise. Facing an avalanche of due diligence requests from investors, the team built an investor relations Agent using ChatGPT.

This Agent was trained on internal data and calibrated to mirror the tone of a public-company-grade investor relations professional. It could deliver data-backed, factually accurate, consistently messaged, and strategically framed answers within minutes. More importantly, since there's only one CFO, this Agent enabled everyone on the team to provide "CFO-level" responses.

The result: both fundraises were completed entirely in-house, saving hundreds of millions of dollars in advisory fees. The Agent was also shared with the recruiting team to help explain equity value to executive candidates.

ChatGPT for Excel: Building an LBO Model in 10 Minutes

Stacie drew on her own experience as a former private equity analyst pulling all-nighters to build leveraged buyout (LBO) models, showcasing the power of ChatGPT for Excel.

The workflow is remarkably simple: upload an equity research report PDF, and instruct it to "think like an investment professional, organize the financial projections, and build an LBO model." ChatGPT for Excel automatically plans the analysis structure, populates assumptions, builds detailed capital structure and cost-of-capital assumptions, even includes a sources and uses table, and finishes with an investment recommendation.

The entire process takes about 10 minutes, and all formulas are written directly into the workbook — fully traceable, auditable, and adjustable for scenario analysis.

Codex: Giving Non-Technical Staff Programming Capabilities

Codex empowers finance professionals without technical backgrounds to build software tools, primarily in three areas:

Marketing ROI Dashboard: Upload massive marketing datasets into Codex, and it automatically generates an ROI dashboard where you can toggle between channels to view spend and diminishing return points. Based on this, the team now rebalances marketing spend weekly, shifting budget from underperforming channels to high-performing ones.

Sales Insights Dashboard: Codex pulls interaction data directly from Gong call recordings and customer emails, displaying by segment, region, and account level which sales reps are pitching new products to customers — no need for salespeople to manually enter data into the CRM.

Automated Executive Reporting: The monthly compute margin slides presented to the CFO used to require coordinating raw infrastructure telemetry data across products and GPU types, then layering on accounting rules. Now, through Codex and reusable skills, days of work are compressed into hours.

Organization-Level Agents: Four Automation Scenarios

At the organizational level, OpenAI's finance team has built Agents embedded into workflows that automatically handle large volumes of tedious routine work:

1. Procurement Agent: Automatically handles approximately 60% of procurement and travel policy inquiries, with continuous improvement
2. Credit Review Agent: Previously required credit risk analysts to research each customer individually; now the Agent completes assessments in minutes, with scores embedded directly into the CRM system
3. Contract Review Agent: Ingests agreements in bulk, automatically flagging non-standard terms (such as clauses affecting ASC 606 revenue recognition), enabling agreement volume to grow without linearly scaling the accounting team
4. Vendor Risk Agent: Automatically completes risk research and scoring, embedded into the approval workflow of procurement software

Key Takeaway: Building an AI-First Management Culture

Stacie highlighted two critical success factors:

AI-First Mindset: Before tackling any challenging task, the team has developed a habitual reflex — "What can ChatGPT do for me?" or "How can I use AI to make this task simpler?"

Democratized Access: Through hackathons and similar initiatives, ChatGPT and Codex are put into the hands of the people closest to the problems, most familiar with the data, and working in the systems every day. Stacie noted that you'd be surprised by the innovations they produce.

Conclusion

OpenAI's finance team proves that an AI-native organization isn't a future concept — it's a present reality. The key isn't about using the most advanced models; it's about transformation on three levels: restructuring the organization (embedding engineers in business teams), shifting the culture (AI-first thinking), and having the courage to iterate continuously (acting before perfection).

20% of the headcount delivering 100% of the work — that number alone is the most powerful argument. For finance teams exploring AI transformation, OpenAI's experience offers a clear path forward: start by redesigning processes from scratch, rather than simply layering tools on top of old workflows.