Against the doomsayers on AI

Whilst we wait for artificial intelligence to inexorably turn our lives upside down, rapidly destroying jobs, inevitably making us all more vulnerable, inescapably messing with our minds and irrevocably creating a bleaker future for everyone – whilst we wait for all this to happen, and it certainly will – there are some counterintuitive and even surprising truths about AI and its everyday applications that seem tailor-made to make life harder for the prophets of doom – and perhaps not just for them. The most distressing debate on AI, as we know, does not centre solely on the possibility that AI might be capable of creating viruses that threaten our future – as Walter Veltroni grimly reminded us yesterday in "Corriere della Sera" – but centres, first and foremost, on the systemic risk of finding ourselves in a world where artificial intelligence will inexorably take the place of human beings. In 2025, as we know, artificial intelligence in the United States experienced a boom, yet despite this, America continued to record more than five million gross hires per month and saw no surge in the so-called aggregate redundancy rate. It was not an employment boom: recruitment figures fell compared with the previous year. Nor, however, did the exodus from offices predicted by the prophets of doom materialise. In the nearly four years of the generative AI revolution, since the launch of ChatGPT in November 2022, the US economy has added around 4.6 million jobs to its payrolls. And of the job cuts announced by US companies in 2025, only 4.5 per cent were explicitly attributed to AI. In 2026, the number of cuts attributed to AI increased, but so far there has been no general surge in redundancies.

Will it last? Who knows.

Among the myths debunked – at least for the time being – by artificial intelligence is another significant one, highlighted in recent weeks by the Wall Street Journal. Not only has AI not yet brought about the ‘job apocalypse’, but with growing awareness of its potential, a surprising selective acceleration in recruitment is emerging. Some of the largest American companies, including Alphabet, Google’s parent company, and the rail giant CSX, have informed investors in recent months that they intend to take on more staff to meet their growth targets and capitalise on emerging technologies (IBM has announced that by 2026 it will triple ‘entry-level’ recruitment in the United States). Artificial intelligence, as we know, introduces a great deal of uncertainty for employers: as the Wall Street Journal also notes, no one knows whether more or fewer people will be needed in the future. But what is currently becoming apparent is that the rise of technology is prompting companies to recruit people who understand how to make it work, how to make the most of it, and how to turn it into a tool capable of creating greater opportunities and higher profits.

An interesting finding on this topic comes from PwC’s Global AI Jobs Barometer 2026: compared with the 2018 baseline, companies in the sectors most exposed to AI recorded a 52 per cent increase in staff numbers, compared with 36 per cent for less exposed companies. And over the same period – another stark statistic for the doomsayers – salaries rose by 24 per cent and 17 per cent respectively. Not bad at all. Within this picture, there is also an interesting detail for those who work with AI. David Brooks noted this a few weeks ago in "The Atlantic": anyone who uses AI for work knows exactly what Brooks is talking about. Remember when it was said that artificial intelligence would offer the chance to have more free time? So far, it seems, that hasn’t been the case. Researchers at ActivTrak analysed the digital activity of over ten thousand workers and found that, with the adoption of AI, working life has become more intense, not less so. Researchers at the Haas School of Business at the University of California, Berkeley, as Brooks reports, have found that, with the use of artificial intelligence, workers have started to take on tasks they previously delegated; they have squeezed in short bursts of work in the evenings, at weekends, in waiting rooms and whenever they had a spare moment and the AI was to hand. AI has not simply made the old work faster: it has expanded the amount of work that people feel capable of and willing to do.

AI was expected to trigger widespread unemployment, and, fortunately, it has not done so yet. It was expected to reduce employment opportunities, and, fortunately, that has not occurred yet. It was expected to grant users more leisure time, and, unfortunately, it has not done so consistently.

Sam Altman, CEO of OpenAI, admitted as much a few days ago. On Ti Morse’s ‘Relentless’ podcast, published on 25 July, Altman said that technology has long promised to make us work less, but that the four-hour working week, to name but one example, has never materialised on a large scale and is unlikely to take hold. His reasoning is simple: when productivity increases, we are not content simply to produce the same amount in less time. The mechanism, according to Altman, is different: expectations rise, new needs are invented, and we feel compelled to do more. This productivity gain therefore fuels a new form of competition, and Altman’s – dramatic! – prediction is that, even in the world of superintelligence, ‘we’ll all be much busier’. ActivTrak, an American software company, recently analysed 443 million hours of activity across 163,638 employees and ran some calculations. In a sub-sample of 10,584 users, monitored for 180 days before and after the adoption of AI, the time spent in all measured categories increased: email by 104 per cent, chat and messaging by 145 per cent, and management software by 94 per cent. The third counterintuitive finding to unpack during this summer of AI-induced panic concerns another little-explored aspect. In public debate, discussions about AI often focus on the risks of overusing it. What is unfortunately overlooked, however, is a symmetrical point: the risk lies not only in using AI too much, but in using it poorly and ending up wasting time rather than saving it. A paper by the ILO (International Labour Organisation), published on 6 May and much discussed amongst experts, highlighted an interesting paradox: AI significantly boosts productivity in individual tasks – generally by between 10 and 70 per cent – particularly in well-defined, text-based tasks and for less experienced workers. However, the benefit does not automatically translate from individual tasks to business outcomes: adoption is uneven, concentrated in large, digitalised companies, and often remains confined to pilot projects. In short: AI is being used more and more – a quantitative fact – but not all companies are yet able to translate its use into productivity – a qualitative fact. In this summer of AI-fuelled panic, therefore, there are some counterintuitive figures that should give us pause for thought, steering us away from dystopian scenarios and bringing us closer to reality. Productivity gains in individual tasks are already evident, but at company level they remain uneven, and overall there is still no sign of a rise in productivity clearly attributable to AI. The predicted employment apocalypse, however, has not materialised – at least not yet. And the potential of artificial intelligence has, in an unexpected way, created a need for talent, for ‘mini-coaches’, for technical directors capable of knowing how to deploy algorithms. The prophets of doom will therefore, for the time being, have to redirect their affectionate thoughts towards other areas. AI was supposed to cause mass unemployment, and so far, this has not happened. It was supposed to make natural talent less appealing than artificial talent, yet so far it seems to have made human talent – the kind capable of harnessing it – even more valuable. Artificial intelligence will change the world, and we know this, but rather than focusing our attention on the issue of rules – on how to cage it, how to stop it, how to regulate it – it might be worth making a little extra effort and asking ourselves, in this summer of AI panic, what can governments and companies do to ensure they have an ever-increasing number of talented individuals capable of using artificial intelligence to combat the universal doomsday rhetoric promoted with natural indolence by the new and old prophets of doom.