AI at work is gradually establishing itself in companies as a tool supposed to solve the problem of productivity. Governments believe in it too. In March 2026, British Chancellor Rachel Reeves announced £2.5 billion, or around €3 billion, of investment in AI and quantum computing to boost the country’s economic growth. Several recent research projects, however, invite us to qualify this enthusiasm and take a closer look at what technology really does, and for whom.
AI at work cuts the time needed for certain tasks by three or four…
AI at work produces considerable time savings on well-defined tasks. According to an article published by Harvard Business School Working Knowledge, a study of 78 employees at IG Group, an international financial broker, measured the impact of generative AI on financial article writing. The conceptualization phase, i.e. structuring ideas and preliminary research, went from 63 minutes on average to just 23 minutes with AI. The writing phase went from 87 minutes to 22 minutes.
These gains apply to all employee profiles, whether they are experts or beginners in the task concerned. Additionally, AI helps particularly well on abstract and structured tasks, like generating ideas or organizing a plan. For this type of work, all participants achieved similar results, regardless of their original specialty.
… but hits a wall when faced with tasks that require real expertise
However, these gains stop as soon as the task requires concrete expertise and accumulated experience. In the study, technical specialists like IT developers scored 13% lower than content experts, despite using AI the same way. The reason is what researchers call “knowledge distance”, that is to say the distance between an employee’s skills and those required by the task. The larger this gap, the less successful the AI is in filling it.
The researchers summarize this phenomenon with a powerful formula. AI can provide the map, but navigating the terrain is another matter. An employee without the necessary foundations will not be able to exploit the AI’s suggestions with sufficient discernment to produce a quality result. Thus, AI amplifies existing skills much more than it creates them.
… and reveals the limits of an economy obsessed with productivity
Beyond the figures, some researchers question the very objective of this race for productivity. According to an article published by Phys.org, researcher Abigail Marks points out that many essential sectors, such as care, education or health, rely on human interaction and cannot be accelerated without degrading their quality. Increasing productivity in these areas often means reducing their real value.
On the other hand, in sectors where tasks are standardized and automatable, AI clearly demonstrates its potential. Therefore, the true measure of AI’s success at work is not reduced to time savings on a spreadsheet. It depends on the ability of organizations to deploy technology where it creates real value, without sacrificing the human quality that makes the difference in the professions that matter most.
