Across several economies, a wave of corporate reversals has followed the initial enthusiasm for AI-driven layoffs. Nearly two-fifths of executives initiated cuts specifically to deploy automation, yet 55 per cent now deem those decisions erroneous, according to recent surveys. Ford has rehired engineers to fix quality problems its algorithmic systems failed to solve, while Australia’s Commonwealth Bank rescinded plans to replace call-centre staff after an AI voice tool caused a backlog of unprocessed calls. The operational disruptions have prompted a re-evaluation of the technology’s readiness for high-stakes, non-routine tasks.

The reversals coincide with a deeper reassessment of AI’s real impact on entry-level hiring. Researchers at the London School of Economics and the University of Warwick found that remote work, not generative AI, best explains the 29 per cent drop in the share of entry-level recruits in several countries—the decline began before tools like ChatGPT were widely available. Meanwhile, a Harvard Business School study of venture-backed startups shows that AI-native firms are 25 per cent smaller, employ roughly 15 per cent fewer entry-level workers and managers, and have a 20 per cent higher share of senior staff. These patterns suggest that AI concentrates expertise rather than democratising opportunity, even as Goldman Sachs economists argue that the technology will displace, not permanently destroy, about 9 per cent of the US workforce.

Viewed from Beijing, a different philosophy is emerging. Chinese firms, constrained by US export controls on advanced chips, have built efficient AI models and humanoid robots that match Western performance with far less computing power. Agibot recently live-streamed eight humanoid robots completing over 64,000 tasks with a 99.99 per cent success rate on a tablet production line, while Zhipu AI’s model rivals top US systems in cybersecurity at lower cost. This scarcity-driven efficiency stands in contrast to the capital-intensive, often hasty deployments in Western firms. In creative sectors, the tension is equally acute: Martin Scorsese’s adoption of generative AI for storyboarding has drawn fierce union criticism, and court battles over whether training on copyrighted works constitutes infringement are pending in the US and EU.

The reassessment is set against a backdrop of cooling labour markets and ambitious capital-raising plans. The US economy added just 57,000 jobs in June, half of expectations, while Goldman Sachs’s CEO described markets as tilted toward “greed” ahead of massive IPOs from OpenAI and Anthropic. Meanwhile, the EU AI Act’s transparency requirements for training data and the outcome of US copyright litigation will shape the next phase. The milestone to watch is whether upcoming AI company flotations sustain the investment case for automation or sharpen the focus on augmentation and oversight.