The 44 jobs most likely to be replaced by AI - is YOURS at risk?

"The test had human experts evaluate tasks carried out by AI and a real professional from that field, and decide which was best. The best performing AI, Anthropic's Claude Opus 4.1, beat human experts 47.6 per cent of the time on average, and had an even higher win rate against a few jobs. When competing against human counter and rental clerks, the AI outperformed the professional in 81 per cent of cases."

"OpenAI says: 'We found that today's best frontier models are already approaching the quality of work produced by industry experts.' Each of the 44 professions was assessed against AI on a set of specific tasks that someone in that role might need to do. For example, registered nurses were asked to assess skin lesion images and create a consultation report, while manufacturing engineers were tasked with making a 3D model of a cable reel."

"The human assessors didn't know which of the tasks was completed by a human and which by an AI; they simply had to choose which they thought was best. The AI's wins were then tallied and averaged into a 'win rate', which measures how well the chatbot performs against real human professionals. OpenAI admits that this doesn't capture the full range of tasks required for most jobs, noting that 'most jobs are more than just a collection of tasks that can be written down'."

OpenAI conducted a benchmark called GDPval comparing AI models to professionals across the nine most financially important US industries and 44 occupations. Human experts performed blind evaluations of tasks completed by AI and by human professionals and selected the better result. Anthropic's Claude Opus 4.1 topped the test, beating human experts 47.6% on average and outperforming counter and rental clerks 81% of the time. Each occupation was assessed with role-specific tasks such as nurses evaluating skin lesions and engineers creating 3D models. Retail and sales emerged as the sector most frequently outperformed by AI. OpenAI notes the benchmark does not capture every job facet but treats it as indicative of likely real-world impact.

Read at Mail Online