The impact of artificial intelligence on productivity: analysis of available research

Other research from a few months ago is also positive. As tools using advances in natural language processing enter business and society, they could result in a 7% (or nearly $7 trillion) increase in global GDP according to Goldman Sachs.

McKinsey estimates that genetic artificial intelligence could add between USD 2.6 and 4.4 trillion per year to the economy. That is an impact of 15-40%. In technology, media and telecommunications, new use cases for artificial intelligence are expected to generate an impact of between $380 billion and $690 billion: $60 billion to $100 billion in telecommunications, $80 billion to $130 billion in media, and $240 billion to $460 billion in high-tech. "In fact, it seems possible that within the next three years everything not related to AI will be considered obsolete or ineffective," writes McKinsey.

Let's come to Europe: one third of European companies have adopted AI, a 32% growth rate over last year that, if maintained, could contribute €600 billion in gross value added to the European economy by 2030 according to a recent Strand Partners report commissioned by Amazon Web Services. This figure, if realised, would bring the total estimated economic impact of technology adoption in Europe to EUR 3.4 trillion by 2030.

In Italy, generative artificial intelligence alone could generate an increase of up to EUR 312 billion in the annual added value of the Italian economy over the next 15 years, equivalent to a growth of up to 18.2 per cent of GDP, according to research by the September TEHA Group (The European House Ambrosetti) in collaboration with Microsoft Italia.

The most famous sceptic, however, is Jim Covello, head of equity research at Goldman Sachs. Three months ago he surprised many with research questioning whether companies would see a sufficient return on their AI spending. According to Covello, generative artificial intelligence, which can summarise text and write software code, makes so many mistakes that it is doubtful it will ever reliably solve complex problems. And it will take years for there to be a positive impact on the economy and productivity.

The well-known MIT economist Daron Acemoglu (in a study from 2024) also puts on the brakes: according to him, the available data, combined with the economic theory of aggregation, support only a moderate impact on total factor productivity and GDP growth, in the order of 0.1 per cent per year.

Part of the economic impact is precisely due to an increase, assumed by many, in labour productivity.

As is well known, income and welfare gains in advanced economies have been held back by weak productivity in recent decades. The growth rate of labour productivity has fallen in OECD economies from around 2% per year between the 1970s and 1990s to 1% in the 2000s. This is a big problem in the face of an ageing population and makes the allocation of resources for the green transition more difficult. AI is seen as the main hope to unlock productivity.

A series of studies conducted over the past year show how generative artificial intelligence can increase the productivity of people in various jobs. Economists at Stanford and MIT found that call centre workers are 14% more productive when using AI conversational assistance; in particular, there was a 35% improvement in the performance of inexperienced and low-skilled workers. Various studies have shown that software engineers can code twice as fast with the help of technology.

Last year, Goldman Sachs calculated that generative AI could boost overall productivity growth by 1.5 percentage points each year in developed countries and increase global GDP by $7 trillion in 10 years. And some predict that the effects will soon manifest themselves.

Anton Korinek, an economist at the University of Virginia, argues that the additional growth has not yet manifested itself in productivity numbers because generative AI needs time to spread through the economy. But he predicts a 1-1.5 per cent increase in US productivity by next year.

According to McKinsey, up to 50 per cent of tasks can be automated in call centres, with the potential to increase productivity by between 30 and 45 per cent, while improving customer experience and satisfaction. As far as the workforce is concerned, up to 70% of repetitive work tasks could be automated through artificial intelligence to improve productivity. There is also potential for new efficiencies in research, validation and knowledge synthesis, where around 60 per cent of tasks have potential for automation. Artificial intelligence tools could increase developer productivity by 20-45%.

Imf estimates that nearly 40 per cent of global employment is exposed to AI. Historically, automation and information technology tend to influence routine activities, but one of the distinguishing aspects of AI is its ability to influence high-skilled jobs. As a result, advanced economies face greater AI-related risks, but also greater opportunities to exploit its benefits, than emerging and developing economies. In advanced economies, about 60 per cent of jobs may be affected by AI. About half of the jobs exposed can benefit from AI integration, increasing productivity. For the other half, AI applications could perform key tasks currently performed by humans, which could reduce labour demand, leading to lower wages and hiring. In the most extreme cases, some of these jobs could disappear. The risk of inequality is highlighted by Imf and the OECD.

Much more cautious is Acemoglu, according to whom productivity growth will only be 0.6 per cent in total over the next ten years.

Differences in estimates depend on how well these technologies will actually improve the way work is done and how much they will be adopted. Adoption certainly depends on the perceived quality of these tools but also on the ability of companies to innovate processes and develop the right skills.

These are all uncertain and depend on many factors: the evolution of AI models and macroeconomic policies of individual countries and Europe, for example.

Optimism about these elements is generally present in studies, with some notable exceptions, however. Even many of the optimists, however, note the risk of increasing inequalities - if the transformation is not well managed by social and economic policies - and this in itself can reflect negatively on economic forecasts.