By Dalton Gardimam, Chief Economist, Ágora Investimentos
Few topics have been as important and potentially disruptive recently as the emergence and consequent impacts of artificial intelligence (AI) on economies, societies, markets and so on. This should come as no surprise, given we’re in the early stages of the process (ChatGPT was introduced on November 30, 2022). Professional economists in academia can already be sorted into “teams”, varying from those deemed optimistic about AI’s evolution in fostering economic growth, employment and productivity on a systemic basis to the skeptical, who essentially believe that AI will make job creation much more difficult with doubtful impacts on or improvements in productivity. Given its persistent low productivity over decades (and disappointing growth) well before AI became a topic of interest, Latin America might find itself in a position to benefit from the process (a fast catch-up?) or, given some of its previous weaknesses, continue to suffer from low productivity (see Figure 1) and persist in delivering sub-par economic growth.
The logic of concentrating on the interconnection of these two major topics—(a) AI’s impact on productivity and (b) Latin America’s lagging productivity—is to highlight any possible opportunities (and associated challenges) for countries in the region due to a new major tool (AI) becoming increasingly available. Many of the inferences here are speculative because today’s “AI” and what it can accomplish now will be vastly, if not completely, different in the future—including the possibility of artificial general intelligence (AGI) (a form of artificial intelligence that surpasses human cognitive capabilities across all ranges of tasks). No matter how fascinating this topic might be, our goal is to narrow the scope of the possible impacts of the technology as we know it today on a region that has systematically fallen short in growth and productivity.
It is important to stress the central role of low productivity in Latin America over the past decades as a key explanatory factor for the region’s sluggish economic growth. It is widely documented in the literature that its low-growth history (relative to its peers) has not been the result of insufficient factor accumulation (capital and labor) but of low productivity. Endogenous growth theory posits that countries with lower productivity can boost it more rapidly by adopting technologies from more advanced economies, reaping the benefits of frontier advances without bearing high R&D (research and development) costs, trial and error, time and so forth. While total factor productivity (TFP) encompasses more than just pure technology—for instance, it includes market inefficiencies, regulatory bodies and economies closed to foreign competition—the catch-up argument equally applies to policies and institutions as well as technologies. Less-developed countries can improve by learning—“copying”—from others without having to start from zero.
The “low-hanging fruit”: Can AI help bridge the gap faster?
Here is the key argument: Latin America, a region with a huge productivity gap and productivity-growth gap, could do more faster and with lower costs to reduce these gaps with a new technology (AI) available to all. The intuitive question is, then: Can AI really help these countries close the productivity gap faster? The rationale behind this is the “low-hanging fruit” concept: Admittedly, it is easier to improve one productive process with many inefficiencies than a state-of-the-art process requiring tiny improvements. In this case, an easy-to-use and relatively cheap but powerful tool such as AI and its different possible use cases can make consequential impacts in less-developed countries.
For that to happen, we have to be absolutely convinced that AI is a finished revolution with a clear-cut known impact on productivity. This assertion is far from a consensus. This is where some famous “skeptical” economists enter the ring. Nobel Laureate Daron Acemoglu (who considers himself an AI realist, not a pessimist) has a dimmer view of AI’s reach over the coming years. In his paper “The simple macroeconomics of AI” (Economic Policy, August 2024), he estimated that AI would produce an increase in gross domestic product (GDP) of between 1.1 percent and 1.6 percent in 10 years—in other words, a 0.05-percent gain in annual terms in productivity. A crucial part of this argument is that only a few jobs (20 percent) will be directly exposed to AI’s impacts.
In the case of Latin America’s productivity gap, it remains wide. However, discussions about some of the reasons for its persistence may link to AI’s potential help in the years to come. The historical reliance on commodities (oil, minerals, agriculture) rather than higher-value manufacturing or services implies a more complicated path for exploiting AI’s benefits, as it is hardly imaginable that changing the economic structure is doable in the foreseeable future. Latin America’s dependence on commodities and, in extreme cases, the “resource curse” or paradox of plenty is a given fact for some of the countries in the region. AI, AI-enhanced systems and AI-designed and -operated machinery can, hypothetically, improve productivity on a sectoral basis (commodity producers becoming more productive, for instance). However, assuming a leap forward in systemic productivity due to widespread AI usage in the primary sector seems a bit exaggerated.
Assuming that the region would start to compete in sectors that are currently not competitive due to AI usage is not realistic. Leaders will likely be in the best position to benefit most from AI tools. No one can dream of Latin America giving birth to a hyperscaler (such as the Big Five: Alphabet, Amazon, Apple, Meta Platforms and Microsoft). Low investments in R&D and innovation in general limit productivity gains. Coupled with addressing human-capital gaps (quality of education, technical skills and training), it seems plausible to envisage an avenue of opportunity from the adoption of AI tools and policies. That is probably where the major gap is situated today and, possibly, where the impacts of AI usage can make a difference (low-hanging fruit).
As the world has started to learn, these achievements and AI’s effective benefits require minimum skills (dubbed as “prompt engineering”, to mention one), with most of them inescapably linked to a good educational system, notably in STEM (science, technology, engineering and mathematics). On this front, the foundational problems of the region’s educational system are an impediment. We can go back to the quality of (basic) education, which explains, at least in part, the gap in human capital later on in the chain. But increasingly, the shortage of minimally trained people to use AI tools is a constraint. This constraint seems to be ultimately parked in the human-capital structure. There should be no doubt that the region still has a long road to improved education (PISA [Programme for International Student Assessment] results are a simple example of that gap relative to Asian peers, for instance). It is a widely established fact that there is a large gap in Latin American education metrics, such as PISA evaluations.
Regarding the bureaucracy, corruption and weak rule of law that have traditionally discouraged investment, increased its cost, reduced profitability and impeded business growth, it would be too demanding to make a connection to AI at this point. Every technological tool can serve as a powerful aid, but improving institutions is a challenge requiring deep and specific efforts in the region. The same can be said about avoiding careless fiscal deficits and inappropriate levels of public debt; this is a basic, ancient, fundamental attribute, being almost an end per se.
Conclusion
The advent of artificial intelligence brings both promise and skepticism regarding its potential impacts on economic growth and productivity. While optimistic economists envision AI as a catalyst for growth, skeptics argue that the gains may be limited and could exacerbate existing challenges, such as job displacement. Perhaps the greatest fear is how to live with some of the possible “positive” impacts on productivity and the resulting job destruction in some sectors. Latin America, with its historically low productivity levels, stands at a crossroads. AI could either be a potent tool for bridging the productivity gap or a missed opportunity if systemic issues persist.
The region’s reliance on commodities, low investment in R&D and gaps in human capital highlight the hurdles it faces to benefit from AI adoption fully. While AI offers the allure of low-hanging fruit in improving efficiency, its true impact hinges on addressing these fundamental challenges. For Latin America to leverage AI effectively, significant strides in education, governance and innovation are essential. Here lies, possibly, the most important line of action for the region: Regardless of whether AI optimists are right or skeptics are more on the spot, Latin America requires an urgent, strong and consolidated effort to improve the educational system at all levels. Only then can the region hope to turn AI from a mere technological advancement into a transformative force for economic growth, enjoying the full benefits of the synergies that would eventually come, producing improvements in productivity and economic growth.