By Arooba Younas The surge in the use of Artificial Intelligence by businesses, corporations, and organisations has been illustrated by the latest McKinsey Global Survey on the state of AI. Accordingly, 88 per cent of organisations reported the regular use of AI in at least one business function, a 10 per cent increase from the results of the survey conducted in the preceding year. Indicative of business leaders’ desire to boost productivity and reduce operational costs, companies adopted AI to achieve greater efficiency. The reverberations of AI integration are felt in the splintering of the organisational pyramid, driven by the expansion of what a single individual can accomplish. However, flattening hierarchies is not tantamount to flattening skill hierarchies or democratisation of expertise. Research has shown that Generative AI (GenAI) has the greatest impact on the level of expertise required to perform jobs, resulting in an accelerated learning curve for entry-level workers. Whilst shortening the learning curve and the time it takes novices to gain competence at new tasks, the limits of GenAI in upskilling workers have been revealed in an experiment conducted by researchers from Harvard Business School’s Data Design Institute and Stanford University, involving 78 employees from IG Group, a UK-based fintech company. Dividing the employees into three groups of occupational insiders (professional writers), adjacent outsiders (marketing specialists), and distant outsiders (technologists), the experiment attempted to gauge whether marketing specialists and technologists could match the performance of professional writers in conceptualising and drafting articles for IG Group’s website while receiving GenAI assistance. The results revealed that AI levelled the playing field at the conceptualisation stage. However, when the nature of the task shifted to executing the ideas by putting them in writing, a prominent expertise gap emerged. Both with and without GenAI’s help, the writers produced the best work, with marketers coming close and technologists lagging behind regardless of AI use. The divergence in performance, despite AI support, is a portrayal of the “GenAI Wall”, depicting the point at which GenAI does not sufficiently reduce the expertise gap between occupational insiders and outsiders. This is indicative of GenAI’s limit in how far it can carry workers outside their domain expertise. Without foundational knowledge, it becomes difficult for outsiders to distinguish between effective and flawed suggestions generated by GenAI. As a result, they may rely on its outputs without applying the necessary refinement, critical judgement and contextual understanding. Nonetheless, the experiment has implications for business organisations, particularly for leaders who overestimate the abilities of AI. It can only hasten the competency of individuals who are adjacent to a domain, and not conjure expertise out of thin air. In this regard, work structure, which is already undergoing change, can be rethought along the lines of expertise. A more fluid work environment can be created, where, in lieu of strict job roles, boundaries can blur by converging adjacent specialists. However, the experiment is not an advocacy for hiring fewer novices. Studies have shown a reduction in entry-level role postings, falling by 32 per cent in 2025 compared to 2019-2022. This suggests a myopic conclusion drawn from GenAI’s putative capabilities, for it is beginners who, when assimilated into the company’s culture and learn through practice, become experts. By thwarting the pipeline of learners by letting AI take over early-career work, which actually led to the making of experts, today’s efficiency may be sacrificed for impoverished talent tomorrow. The latter has been represented by the AI boomerang effect – recent AI-related layoffs, which are now being quietly reversed – as reports indicated employers regretting layoff decisions. This is because of AI hallucinations as well as its inadequacy to replace the required context or institutional knowledge, which human employees possessed. Although the IG Group experiment was conducted in a Western context, its implications can be generalised across the world, particularly as the allure of AI has diffused across continents. However, the embracing of AI in the Global South, whether in businesses or other sectors, has to be accompanied by digital literacy, localised governance, and a research ecosystem. This is significant lest the adoption of AI brings with it downsides in the form of overdependence and a putative magic bullet solution to solving complex issues ranging from bad governance to climate-related disasters. Additionally, GenAI’s exposure to the labour market has caused faster materialisation of disruptions rather than productivity gains due to a higher risk of job losses, as AI-driven automation can close clerical and administrative positions, which have historically served as pathways to decent work. Conclusively, GenAI has brought the promise of enabling individuals to bridge their capability gaps. However, AI cannot replace expertise or be an all-encompassing fix because it relies on pattern recognition rather than understanding real-world context. Thus, the future of work is where humans can learn to effectively partner with AI, grounded in human judgement, and not replacing it. Arooba Younas is a Research Assistant at the Centre for Aerospace and Security Studies (CASS), Lahore. She can be reached at info@casslhr.com.
