When OpenAI finance executive Kyle Kober set out to help automate tedious workflows in his department, the project took him somewhere unusual for someone with his background: developing code.
A key focus of the effort was streamlining reporting and analysis around OpenAI’s computing capacity costs. Kober said he and his colleagues used Codex, an AI coding agent developed by OpenAI, to automate much of the process, cutting from days to hours the time needed to complete one of the most cumbersome aspects of monthly-close reporting within the company.
The project reflects an emerging shift in corporate finance: AI coding technology is giving finance teams a new way to build targeted custom tools without waiting for help from IT or engineering professionals.
“There’s now the opportunity to just do it yourself,” Kober told CFO Dive in an interview.
Before joining OpenAI, Kober led product and corporate finance at Nextdoor. Prior to that, he was an associate at Financial Technology Partners, an investment banking firm focused on the financial technology sector.
Kober now serves as director of product finance at OpenAI, where he focuses on consolidating “product-level forecasts across user metrics, revenue, and compute demand,” according to his LinkedIn profile.
OpenAI’s sweeping finance rethink
The transformation project he worked on highlights a broader push across OpenAI’s finance organization.
In a blog post last month, OpenAI CFO Sarah Friar outlined her effort to build an “AI-native finance function,” which involves moving toward a “zero-day close” as well as automated, continuously updated forecasting.
Kober was a central player in rebuilding some of the team’s manual processes, he said. The computing-related workflow was particularly difficult because of the labor-intensive work involved: Reconciling product-usage data with accounting records, then turning that into analysis and reporting.
With Codex, the finance team was able to simplify the workflow. The team began using Codex after it caught on among OpenAI’s engineers, Kober said. “I think most engineers had their Codex moment in like November or December of last year; it was only a couple of months later for us,” he said.
As a result of the automation push, the time required to complete the compute analysis process was cut from about five days to roughly five hours, Kober said.
AI coding gathers momentum
The changes at OpenAI come as AI-driven coding rapidly gains ground, particularly among software engineers. Last year, Gartner predicted that 90% of enterprise software engineers will use AI code assistants by 2028, up from less than 14% in early 2024.
The trend could also open new possibilities for finance, according to a June report from Boston Consulting Group. AI coding agents can give finance teams a way to build targeted applications for tasks such as analysis, matching and anomaly detection “without waiting for long development queues,” BCG researchers said.
But the productivity gains from AI coding also come with new costs and risks. Gartner has warned that AI coding expenses are rising as token consumption increases and vendors shift toward consumption-based pricing, with the firm predicting that AI coding costs could surpass the average developer’s salary by 2028. Tokens are the basic units of data processed by AI, with greater token consumption generally resulting in higher costs.
Meanwhile, a March paper published by the Cloud Security Alliance, a non-profit dedicated to cybersecurity, said organizations are integrating AI-generated code at scale into production systems despite documented security risks.
BCG cautioned that finance leaders bringing AI coding into their teams need clear guardrails in areas such as auditability.
“Done right, AI-driven development gives finance teams a faster path to the applications they have always needed and rarely had the resources to build,” the BCG report said.