A recent large-scale survey of 9,000 employees by OpenAI showed that its AI tools have helped many employees save 40 to 60 minutes of professional work time every day, especially in positions such as data science, engineering, communication and accounting.Overall, three-quarters of respondents believe AI has improved their work speed or output quality.

Although the AI ​​craze has lasted for three years, there are still doubts about whether it can truly bring productivity dividends. A study by MIT in August this year pointed out that the vast majority of companies have failed to reap returns on their investment in generative AI.

New research from Harvard and Stanford also warns that many employees are actually producing low-quality content known as "workslop" - which appears to be high-quality but is actually not helpful. These conclusions have led to growing warnings about an “AI bubble.”

In response to such doubts, many AI companies began to release data on their own to prove the economic effectiveness of AI. OpenAI competitor Anthropic recently claimed that its Claude tool can reduce the time it takes users to complete tasks by 80%, but such research has not yet been peer-reviewed.

Brad Lightcap, chief operating officer of OpenAI, said the conclusions of the external study were "not entirely consistent" with what the company actually observed.He believes that the speed of enterprise adoption of AI is increasing rapidly, and in some scenarios it is even faster than that of consumers.As of now, more than 1 million companies have paid to use OpenAI's product ChatGPT work suite, and the number of paid seats has reached 7 million.

The survey also shows that employees who use more in-depth, frequently call advanced models and are good at combining multiple AI tools will see greater efficiency improvements. People in engineering, IT and research departments who don't originally write code have seen a 36% increase in coding-related messages sent over the past six months.

Ronny Chatterjee, chief economist at OpenAI, noted: “Three-quarters of workers today say: ‘I can do things that I couldn’t do before.’ This is a point that is often overlooked when discussing AI and work.”