OpenAI finance chief says AI is reshaping finance jobs

OpenAI Chief Financial Officer Sarah Friar said the company is rebuilding its finance function around artificial intelligence, reshaping day-to-day operations across investor relations, tax, and accounting.

The finance team has used internal hackathons, custom GPT tools and newer agent-based systems as part of that effort. Friar said the changes have altered how work is organised and what skills matter for staff and students entering the profession.

She outlined a broad view of how AI is changing finance roles, rejecting both the idea that work will remain untouched and the claim that all jobs will disappear. Instead, she described a labour market in which some routine tasks fall away, many jobs change shape, and new roles emerge.

"I neither believe that AI will have no impact. Don't worry, we'll all be fine. Everything will be steady as she goes. And I also don't believe AI is going to take away all of our jobs," said Sarah Friar, Chief Financial Officer, OpenAI.

OpenAI's own economic research suggests a mixed outcome, she said. Some tasks are likely to move to agents, while other jobs are already changing. Friar added that her own role looks markedly different from a year earlier because of agentic systems.

Finance buildout

OpenAI's finance organisation has about 200 staff, which Friar described as lean for a global business of its scale. At the same time, she said the group is larger than a conventional finance team in some areas because AI creates room for more analytical work.

Those additions include economic research and pricing, functions not always housed under a chief financial officer. At OpenAI, they sit inside finance because the function now has more scope for insight work, she said.

The shift began with practical experimentation. When Friar joined the company about two years ago, the team started with ChatGPT and custom GPTs. She introduced quarterly hackathons and gave staff time away from daily workloads, saying many employees felt too busy to learn new tools even when those tools could later reduce repetitive work.

One early use case focused on fundraising. OpenAI built a custom GPT for investor relations by loading company presentations and diligence material into the system, she said. The tool was used to answer investor questions and was given a defined style and clear limits: stay factual, do not oversell and say when it does not know an answer.

That approach later proved useful in an overseas investor meeting. Friar said she was handed two pages of questions in Korean and used the investor relations GPT on her phone to translate the material and produce answers on the spot. She said the company is close to enabling two-way, direct language exchanges in live meetings.

Routine work

The larger change, Friar said, is in core finance processes. She listed quote-to-cash, order-to-payment, equity, tax, and treasury as the building blocks of a finance function. OpenAI now reviews those areas, task by task, to identify work that software can complete more consistently and at a greater scale.

Accounting and audit work is one example. Traditional audit methods often rely on sampling due to limited staff and time, Friar said. Teams inspect a subset of invoices, but in an AI-based model, an agent can review all invoices rather than a sample. That changes the control environment and the accountant's role, allowing more time to check exceptions and interpret results.

Tax is another area where the company has moved quickly. Friar said the tax team has become one of the strongest internal adopters of OpenAI's tools despite the conservative reputation of tax departments. She described an annual process in which teams must locate forms from tax authorities around the world and fill them out correctly, with even small changes to those forms often creating extra manual work.

At OpenAI, that process is now largely automated, she said. Systems pull the latest forms, recognise the layout and prefill the boxes, and the tax team then checks the results. Friar said this removes a task that few finance professionals are trained for and gives staff more time to focus on judgement and business questions.

Her account points to a broader shift in white-collar work. AI systems are not only handling document retrieval or translation. They are also taking on structured administrative tasks that once consumed large numbers of hours. In finance, where many workflows are rules-based and documented, that change has immediate implications for staffing, controls and training.

Human judgement

Even so, Friar said some parts of finance remain stubbornly human. Fundraising is one of them. AI can improve preparation and response times, she said, but trust still depends on direct relationships.

"In the end, people invest in people," said Friar.

Investment decisions still hinge on integrity and judgement, she said. Money matters, but over time, the people around a company matter more. Investors can offer advice, access and perspective, and that part of fundraising has not changed, Friar said.

That emphasis on trust also shaped her comments on career development. Friar described her own path as indirect: she studied engineering and material science, trained in accountancy, worked in consulting and investment banking, and later moved into operating roles.

Before becoming a finance chief, she said, a headhunter told her she had no idea what a chief financial officer actually did. She took that criticism seriously and joined Salesforce, where she worked under then Chief Financial Officer Graham Smith. That experience gave her the operational grounding she needed before later roles at Square, Nextdoor and OpenAI, she said.

Her advice on status and titles was blunt. Young professionals should not chase roles because they sound impressive or because others approve of them, Friar said. A better guide is to identify the basic skills gained along the way and the work that is both useful and personally satisfying.

Students and skills

Friar framed that advice around three words she said she uses with her own children: curiosity, adaptability and kindness. Curiosity matters because fear often comes from not understanding a tool, she said. Adaptability matters because careers will shift as AI changes the nature of work. Kindness matters because professional relationships last for years and often return in unexpected ways.

"I use three words when I talk about thinking about AI, career, future. Number one is curiosity. Number two is adaptability. And the third I will always say to everyone, which is kindness, which is probably an unexpected third in some ways," said Friar.

Universities should experiment widely with AI rather than hold back, Friar said. Institutions should also redesign teaching and assessment to account for the presence of these tools. In her view, the key question is not whether students will use AI, but whether courses still test real understanding when AI is available.

She said there was early concern that students would use chatbots to cheat, but argued that access to outside help is not new and that educators should focus on assignments that reveal how students think. She gave an example from her daughter's chemistry studies: the student uses chat tools when stuck on a problem instead of waiting for a teaching assistant. Friar said that it can improve learning, provided formal exams still require mastery without assistance.

She also argued against the idea that AI inevitably destroys professions. Friar cited software engineering as an example, noting that the number of software engineers has increased since ChatGPT emerged. She concluded that AI can widen participation in technical work rather than narrow it.

Cross-disciplinary training is likely to matter more, she said. Finance teams will still need technical accounting and tax knowledge. They will also need people who can frame problems clearly, interrogate outputs and work across business lines. That mix reflects the change Friar described within OpenAI's finance group, where repetitive tasks are increasingly handled by software and more human effort shifts toward analysis, review, and decision-making.

"You all are not in school right now to learn how to fill in pdfs right," said Friar.