Anthropic CFO Krishna Rao claims that AI now writes 90% of the company's code — and this is fundamentally changing the daily work of "white-collar" workers.
Anthropic CFO Krishna Rao claims that AI now writes 90% of the company's code — and this is fundamentally changing the daily work of "white-collar" workers.
In an episode of Patrick O'Shaughnessy's podcast "Invest Like the Best," Rao described the workflow inside Anthropic: AI systems are increasingly taking over the level of intellectual tasks, from software development to financial reporting. At the same time, the role of humans is shifting towards oversight, decision-making, and strategic planning, Business Insider writes .
“We hired significantly more people because of this,” Rao said, adding that he sees Claude as a productivity “accelerator.”
“It actually means we can get a lot more done. Even as our team grows, new hires become more productive by learning how to use Claude within the company. And I think this trend is starting to show up in many other organizations.”
These comments provide a glimpse into how one of Silicon Valley’s leading AI companies sees the transformation of white-collar jobs. It’s not necessarily about completely replacing workers, but rather about automating a significant portion of their daily tasks. It’s turning employees into supervisors overseeing the work of AI systems.
According to Rao, such changes are already happening within Anthropic itself.
“Over 90% of our code is actually written by Claude Code,” he stated, referring to the company’s own code writing product.
The company’s finance team is undergoing a similar transformation, Rao said. Anthropic already uses Claude to prepare financial statements, and the monthly financial audit process is “90-95% complete” before humans come in to review and interpret the results.
"What used to take hours," Rao noted of some internal reports, "now takes just 30 minutes."
Rao believes that increased productivity has allowed Anthropic employees to spend less time gathering information and more time making decisions based on it.
“I see this as a way to highlight and accelerate the potential of the talent we already have,” Rao said. “We talk a lot about the quality of talent being more important than the quantity, and I think that’s exactly what’s happening here. We wanted to bring together the strongest core of AI researchers and model optimization engineers. We believe that combining those with the best models is a truly winning combination.”
Rao noted that as workers become more productive with the help of AI, companies can even expand their workforce because “there is always enough work.”
But Anthropic’s description of the workflow suggests that the very nature of intellectual work is already changing rapidly. He says that employees are increasingly not performing tasks manually, but overseeing AI systems while entire “fleets of agents” work on different projects simultaneously.
“Everyone becomes a kind of manager,” says Rao. “The implications of that and the productivity gains we can get — we’re only at the very beginning of this journey, but the potential is just incredible.”
