From chatbots serving citizens and consumers online to automated power grids providing water supply and electricity to communities, artificial intelligence has become ubiquitous, pervading all facets of our work and personal lives.

But the speed with which it has “invaded” is frightening, raising serious concerns about the negative repercussions.

OpenAI, through its GDPval study, attempted to quantify the impact of AI on real-world work by analyzing 44 professions and their tasks, assessing how effectively AI models can already perform these duties at a quality comparable to, or exceeding, that of humans.

The results are both impressive and worrying. Many of the 44 professions studied were found to be at the “threshold” where machines perform just as well, or even better, than humans.

OpenAI researchers used real tasks from various professional fields—including legal opinions, financial analyses, engineering designs, articles, and business plans—and asked advanced models such as GPT-4o, Claude Opus, and GPT-5 to complete them.

The AI outputs were then presented anonymously to professional evaluators, who did not know whether the work had been done by a human or by AI. They were asked a simple question:

In many cases, AI-generated work was indistinguishable from human output, and often, the evaluators judged it to be superior to the average professional’s work—faster, more consistent, and error-free.

Jobs Most at Risk

The study concludes that knowledge-based, repetitive, and measurable jobs are the most vulnerable to automation.

For example, lawyers and accountants saw AI models draft contracts, analyze legislation, and prepare audit reports with remarkable accuracy. Journalists and editors are also affected, as content creation, data analysis, and text generation are now areas where AI performs highly effectively.

Similarly, engineers and technical designers observed AI generating blueprints, running calculations, and proposing solutions in seconds. Even in finance, administration, and customer support, AI-driven automation already produces consistent results once handled by human teams.

However, OpenAI notes that in most cases, AI doesn’t fully replace humans, instead shifting their role toward supervision, reviewing, approving, and correcting machine output.

One of the key and most impressive findings of the study is how rapidly major AI models have adapted and “self-improved” over the past two years to surpass humans in performing tasks once believed immune to automation.

OpenAI acknowledges the study’s limitations: the results focus on isolated tasks and do not account for collaboration, creativity, or moral judgment—areas where humans still maintain an advantage.

Still, the message is clear: AI is no longer just “next to” the worker but within the work itself.
Journalists, lawyers, economists, and engineers—all must now learn to use, oversee, and integrate these systems rather than ignore them.