The advancement of artificial intelligence is causing a major debate among economists, mainly about how to leverage the technological transformation it brings to significantly increase productivity , growth, and income in national economies.

The perception is that the magnitude and distribution of these gains will depend not only on advances at the technological frontier, but also on the capacity for technology diffusion among the sectors of a country's economy, whether it is advanced or emerging.

Based on these elements, economist Solange Srour , director of macroeconomics for Brazil at UBS Global Wealth Management ( UBS GWM ), wrote an article, which NeoFeed had first-hand access to, in which she places Brazil in this debate from the perspective of the long-term stagnation of Brazilian productivity, which she describes as a "well-documented structural weakness".

Titled "Artificial intelligence, diffusion and productivity: the risk of divergence and the Brazilian case ," the article – which included the collaboration of economists Débora Nogueira and Victoria Roquetti from UBS GWM – seeks to analyze how the development of AI can help increase labor productivity in the country.

According to her, the path to Brazil's progress in this area involves replicating the productivity gains of agribusiness , the only consistent example in the Brazilian economy, in the service sector — which concentrates most of the employment and determines the behavior of aggregate productivity, essential to consolidate this growth.

The relationship between AI advancements and increased economic productivity is central to the current macroeconomic debate. In Brazil, this discussion is even more relevant. As the author observes, Brazilian productivity has grown slowly and unevenly since the 1980s, with long periods of stagnation, reflecting persistent deficiencies in human capital, infrastructure, the regulatory environment, competition, and management quality.

The figures presented in the study give a sense of the challenge Brazil faces. Between 1981 and 2024, productivity per hour worked grew by 0.5% per year, while per capita income increased by 1.0% per year, benefiting from a demographic bonus and an increase in the participation rate. Between 2010 and 2024, productivity grew by only 0.3% per year.

"Recent growth in per capita income has been predominantly extensive, that is, based on the incorporation of more people into the labor market, and not on efficiency gains," the author points out.

Between 2019 and 2024, income grew by 1.7% annually, of which 1.1 percentage points resulted from the increase in the employment rate, while productivity accounted for only 0.3 percentage points. "This model, however, has lost its sustainability with the virtual exhaustion of the demographic bonus," he adds.

In this context, Srour states, agriculture stands out as an exception, operating close to the technological frontier and rapidly incorporating artificial intelligence tools.

“Unlike other sectors, agriculture operates in an environment more exposed to international competition, with greater incorporation of technology, integration into global supply chains, and less regulatory protection, factors that favor continuous efficiency gains,” says Srour.

The impressive productivity figures for the sector cited in the article are noteworthy. "Between 1996 and 2024, agricultural productivity grew by about 6% per year. In 2023, it registered an expansion of 22.3%," it says. In 2024, however, it slowed to 1.6%, contributing to aggregate productivity growing by only 0.1%.

Services in focus

The central challenge of technological advancement, the author observes, is not only to expand specific productivity gains, as in agriculture, but to create conditions for digitalization and AI to spread throughout the service sector – which concentrates most of the employment and determines the behavior of aggregate productivity.

"Without this advancement, AI tends to reinforce productive heterogeneity and widen internal inequalities, instead of functioning as a vector of convergence," he warns.

The service sector accounts for approximately 70% of working hours in the country. According to the study, historical evidence shows that only during periods of consistent productivity growth in this segment did aggregate productivity exceed 1% per year.

"This is a sector characterized by high fragmentation, low competition, a strong presence of informality, and a complex regulatory environment, factors that hinder economies of scale, investment in technology, and the dissemination of good management practices," says Srour.

According to her, as a consequence, innovations tend to spread slowly and unevenly. "The growth of the Brazilian economy is, therefore, structurally conditioned by the capacity to increase the efficiency of services."

The author argues that if AI accelerates productivity only in restricted sectors, such as agriculture, the macroeconomic effects will be limited. However, if there is widespread diffusion into services—such as commerce, logistics, healthcare, education, and business services—the impact on potential output could be significant, reducing supply constraints and moderating structural inflationary pressures.

For Brazil to capture this transformation favorably, Srour cites four pillars as fundamental: consistent investment in digital and energy infrastructure; human capital development focused on digital skills, analytical reasoning, and English proficiency; increased access to financing for technological adoption, especially for small and medium-sized enterprises; and strengthening the regulatory framework.

“The experience of Brazilian agriculture shows that when human capital, external integration, productive scale, and economic incentives are aligned, convergence is possible,” says the author. “The challenge now is to replicate this logic in the sectors that determine the aggregate performance of the economy.”

Srour concludes the article by highlighting the need for the country to define a productivity growth model based on advances in AI.

“The strategic choice is not between leading the technological frontier or not, but between allowing technology to concentrate in islands of efficiency or promoting a broad diffusion that transforms the productive base,” she states, noting that AI can act as a vector of convergence or as an amplifier of asymmetries.

"The outcome will depend less on the technology itself and more on the institutional, educational, and macroeconomic decisions adopted in the coming years," he adds.