OpenAI drops economic study on ChatGPT's job impact

For months, people have been debating whether employees are actually using generative AI to work or just to pass the time. OpenAI just showed its hand with some staggering internal metrics.

OpenAI Chief Economist Ronnie Chatterji and his team released the company's first formal economic analysis of ChatGPT. Along with it came updated adoption figures: ChatGPT now has over 500 million active users globally, sending more than 2.5 billion messages every single day, with 330 million daily messages coming from the US alone.

Workplace adoption is surging fast. Among employed US adults who have ever used ChatGPT, 28% now report using it on the job, up from just 8% back in 2023.

Bureaucrats, teachers, and a new DC outpost

The report highlights several concrete time-saving figures across sectors.

State workers in Pennsylvania are reportedly saving an average of 95 minutes per day on routine administrative tasks. Meanwhile, teachers are clawing back nearly six hours a week on lesson prep and grading duties.

To formalize this research, OpenAI is kicking off a 12-month study with prominent economists Jason Furman of Harvard and Michael Strain of the American Enterprise Institute and Georgetown University. The trio plans to build new metrics tracking AI's real-time impact on the workforce and labor productivity.

The project will be based inside OpenAI's newly announced Workshop in Washington, DC, which will also host hands-on demos and training sessions for policymakers and business leaders.

OpenAI compares the technology to the wheel, electricity, and the transistor. But the company openly acknowledged the darker side of rapid adoption, noting that while the broader economic pie will grow, the transition will be jarring as certain jobs disappear and others evolve at unprecedented speed.

Why it matters

Hard usage data beats Silicon Valley hype every time. Moving from 8% to 28% workplace penetration in under two years proves that ChatGPT has graduated from an experimental chatbot into default office software. If you run a business or work in knowledge work, your colleagues are almost certainly using it under the hood whether your company has an official policy or not.

By planting a research flag in Washington DC with heavyweight economists from both sides of the aisle, OpenAI is playing smart politics. Regulators and unions are increasingly vocal about automated layoffs, and OpenAI wants to anchor the public narrative around public sector savings and teacher support before stricter labor rules emerge.

We think the narrative framing is sharp, but it skirts the toughest issue: who captures the value. Saving a government clerk 95 minutes a day sounds heroic, but in private enterprise, compressed work cycles historically lead to higher quotas or reduced headcounts. OpenAI can build the tools to scale human output, but corporate executives, not foundation models, will decide whether those hours turn into layoffs or leisure.