OpenAI warns that artificial intelligence can do more than most people imagine

This gap is not just a technical issue, but has important social and economic implications. The wider the gap, the more likely it is that the benefits of AI will remain limited to a small circle of users, while the majority will only benefit from its superficial part.

How deeply do people really use artificial intelligence?

To better understand this phenomenon, OpenAI measured not just the number of users or frequency of use, but the depth of use. As a proxy, they used the measure “thinking capabilities” (thinking capabilities), which reflects the amount of reasoning and effort the model has to put into answering. The more the user asks complex questions or delegates multi-step tasks, the more these capabilities are activated.

Analysis of anonymized data shows a marked gap between the so-called “power users” and typical users. Users in the top five percent of usage intensity use, on average, about seven times more thinking capacity than the average user. This means that even with access to the same tools, most people use only a fraction of their capabilities.

The gap is actually even wider. According to internal estimates, OpenAI employees use AI significantly more intensively – more than ten times more intensively – than typical subscribers of paid packages. This shows that knowledge and usage play at least as important a role as access to the technology itself.

Country differences: surprises outside the rich world

A similar pattern emerges at the country level. Among the more than 70 countries analyzed, the leading countries use, on average, about three times more thinking capacity per capita than the countries at the tail of the scale.

Some large and wealthy powers are expected to be at the top, but countries such as Vietnam and Pakistan are also among the most advanced users. This shows that a high level of economic development does not in itself guarantee better use of AI. It is often users in developing countries who are quick to use new technologies more boldly and deeply.

From chat to work: where the biggest differences arise

OpenAI finds that the differences between users widen most when AI usage moves beyond basic chat. While simple uses—finding information, creating text or images—are relatively evenly spread, the gap widens dramatically for more complex tasks.

The biggest differences occur in programming and the use of more advanced features such as data analysis, multi-step problem solving, and so-called "agentic tools” (tools that allow AI to independently perform a sequence of tasks). In these cases, leading countries and the most experienced users use AI many times more often than average.

It is also telling that a significant proportion of even paying business users never access more advanced features. This shows that the barrier is not just price or access, but rather understanding how to integrate AI into repeatable work processes.

Untapped potential as a social challenge

OpenAI warns that “capability overhang" will not reduce on its own. If the development of models continues faster than the spread of knowledge and practices, the differences may deepen. In this case, AI would become a tool that reinforces existing inequalities, rather than reducing them.

The key concept here is "agency” – the ability of individuals, organizations, and countries to actively use artificial intelligence to solve concrete problems, not just as a passive tool for quick answers. Access to AI is only the first step; real benefits only arise when people develop the skills to use it effectively.

OpenAI points out in the report that artificial intelligence has the potential to be similar to electricity or the internet: it can greatly increase productivity and living standards. However, the reality is that this potential will not be realized automatically.

Unless knowledge, training, and opportunities for deep use are made more widely available, the benefits of AI will disproportionately accrue to a smaller circle of advanced users and countries. Closing “capability overhang"Bridging the gap is therefore not just a technical task, but one of the key social tasks of the coming decade."