Anthropic CFO says AI now writes 90% of the company’s code
Krishna Rao says Claude is reshaping work inside Anthropic, moving teams from execution-heavy tasks toward supervision, review, and decision-making.
Summary
Krishna Rao, Anthropic’s CFO, said AI now writes more than 90% of the company’s code through Claude Code. He made the comments on Patrick O’Shaughnessy’s “Invest Like the Best” podcast.
According to Rao, the shift is not limited to engineering. Anthropic’s finance team also uses Claude to prepare financial statements and internal reports, with monthly review processes reaching “90 to 95% ready” before humans step in.
In practice
Rao describes a workplace where AI increasingly handles the execution layer of white-collar work. Tasks that used to take hours can now be completed in about 30 minutes, giving employees more time to interpret results and make decisions.
In his view, this changes the role of teams. Instead of manually completing every task, employees supervise AI systems, validate outputs, and direct “fleets of agents” across multiple projects at once.
Context
The comments come as companies across industries debate whether AI will replace skilled workers or make them more productive. Rao argues that, inside Anthropic at least, Claude acts as a productivity accelerant.
He also said Anthropic has hired more people because of these productivity gains, arguing that there is “no shortage of work to do”. At the same time, his description suggests that administrative and technical work is changing quickly.
Why it matters
- Anthropic is showing internally the kind of transformation it sells to the market.
- Automation is no longer limited to simple tasks; it now touches code, finance, and reporting.
- Human value shifts from execution to supervision, validation, judgment, and strategy.
- For companies, the question is no longer just “using AI” but redesigning roles around agents.