Will AI steal white-collar jobs?

As an editorial dealing with "Strategic Trends" familyandtrends had its say onartificial intelligence a year ago, a lot has happened since then but nothing that wasn't predicted. Last week, Citrini Research's report triggered a half-earthquake by describing a hypothetical deep crisis scenario in June 2028 created by the unemployment of large numbers of white-collar workers replaced by 'Ai agents who don't sleep, don't get sick and don't require health insurance'.

The report created the 'excuse' that the financial markets needed to sell the overvalued software companies, which, confirming the Citrini scenario, are already announcing major plans to lay off code writers, and this served as a megaphone for even greater generalised concerns. The Economist intervened by stating that "a research report that went viral got the economic calculations wrong" and after calling out the highest calibre of British economics such as David Ricardo, Thomas Malthus and John Maynard Keynes in support of his thesis, he concluded that if successful AI companies did not reinvest the cash they earned there would be deflation, unemployment and a fall in GDP!

Let us try to summarise what the report claims and then return to the laws of economics.

First. Citrini predicts a vicious cycle in which AI capabilities improve, firms need fewer white-collar workers, more white-collar workers lose their jobs, the unemployed spend less, margins shrink, firms need to save money and invest in more AI, and the cycle continues. As familyandtrends already said, the impact of AI on intellectual labour will be similar to that of the robotization of factories on manual labour: more standardisation of quality, less room for craftsmanship, more capital and less labour, more skilled labour. For those like familyandtrends who are old enough to have lived through it in the industrial Italia of the 1980s, this is nothing pleasant, back then the class struggle came to an end, the 'mega-factories' were emptied and abandoned, and the term 'blue collars' was coined to identify those who footed the bill for robotization. This evolution of the economy will not be simple, it is no coincidence that it was OpenAI in its most idealistic infancy that financed the largest experiment of 'basic income', i.e. citizenship income, to understand the effect of sudden unemployment that is slow to reabsorb: the data is public and the reading is interesting. The fact remains that the social economic system in which we live has outgrown economic developments similar to the one Citrini hypothesises, not least because in his vicious circle he does not delve into the investment part. The report predicts: "the news that GE Vernova's entire turbine production capacity would be exhausted until 2040", and this point is illogical: in reality either the price of turbines or the number of GE's competitors will have increased, and those selling turbines will have more capital invested and/or more workers. There may be more blue-collar workers engaged in turbine production and fewer white-collar workers engaged in administrative tasks in turbine companies, and the former may be paid more (and if so, the champions of reshoring and tariffs will suddenly look smarter); it may also happen much faster than in the 1980s because the investors will not be manufacturing companies with low margins and debt constraints but supertechs with high margins and sitting on mountains of cash: faster does not mean sudden, however.

Secondly, 'SaaS was not dead. There was still a cost-benefit analysis to be considered for the management and support of in-house developments. However, the option to develop in-house existed, and this weighed on price negotiations'. Saas, made possible by bandwidth and server farms, allowed software companies to concentrate production activity in one place and remotely update software to all customers, this generated low-priced standards and customisation complexity, AI will make customisation low-priced and in-house. Again, this will not happen overnight, it will take time to change the infrastructure and bring the necessary skills in-house; how much? Less than the time it took enterprises to abandon the IBM AS400, but not much less.