Who really uses AI? Who benefits the most from it? Well, the answer certainly won’t surprise you. According to a recent study by Perplexity with Harvard University, we are facing a very limited number of professions working in technology-intensive, knowledge-intensive, and data-intensive sectors, notably digital technology, research/education, finance, marketing, and entrepreneurship.
Unsurprisingly, digital professions lead the way:
- It is the largest professional group in terms of adoption and usage.
- It represents 28% of all users adopting the agent and 30% of agentic queries.
- This category notably includes professions such as software development, computer engineering, data & AI, cybersecurity, etc.
They are followed, also not surprisingly, by research professions who make intensive use of it for learning, information synthesis, literature review, and other cognitively demanding tasks.
Next come finance professions who use these tools to filter financial data, synthesize reports, explore opportunities, and automate certain analytical processes.
The field of marketing and sales is not left behind, especially online marketing and sales, which use agents for content creation, trend analysis, campaign optimization, and profile management.
Finally, the tail end of the leading group is occupied by entrepreneurs.
The result of an analysis of hundreds of millions of queries and users
This discovery comes from several months of analyzing data from Perplexity’s Comet browser, one of those AI browsers that allow both web browsing and chatting with an AI. Several million sessions helped determine who used AI and for what purposes.
But beware, not all of these are agentic; only the most complex queries are: those that involve, for example, the agent taking control of the browser, or those that involve multiple external applications to act.
The more people use AI, the more complex things they ask of it
Another interesting insight from this study is how AI usage evolves among these professions.
Initially hesitant, the more familiar they become with the tool, the more complex tasks they will ask of it. This progression showed that AI does not replace skills but allows its users to focus on the most strategic tasks requiring the most intelligence. A finding almost counter-intuitive to what is often heard about AI. Rather than degrading human intelligence, it would serve it.
A more surprising fact is that the use of agents within this very population of early adopters is extremely concentrated. What the study calls « Power users » shows that they make nine times more queries per day than average users, while alone accounting for over 50% of queries.
One could always accuse Perplexity of boasting its own merits, but it is interesting to see that the use of AI agents is truly progressing and that once adopted, users do not revert to simple queries. This leads me to believe that today, those who adopt these agents fastest will have an undeniable competitive advantage in the years to come. AI does not replace them, contrary to popular belief; it makes them much stronger, as the head of Nvidia said, and even gives them more work. This is another fact that confirms my theory that if we don’t pay attention, AI will tend to increase productivity and intelligence gaps in the population.
There will be those who take the bull by the horns, and those who will merely follow without truly grasping what AI can bring them. The problem of acculturation and training therefore remains the central challenge for integrating AI into the business world and for young people.