Innovation Practitioners and Scholars on AI: What Research Can Foster AI-Driven Innovation?

Artificial intelligence (AI) has the potential to become a transformative, general-purpose technology, reshaping global economic, social, and institutional frameworks (Agrawal et al., 2022). It is also remaking the very foundations of research, development, and innovation, to become a general-purpose method of invention, transforming how knowledge is generated, tested, and applied  

This transformation is also changing the underlying economics of innovation. AI is driving a transition toward more capital-intensive innovation models, where data infrastructure and computational capacity are playing a growing role relative to human labour. These changes make AI adoption particularly sensitive to complementary assets, such as computing power, cloud access, digital systems, and proprietary data. In OECD economies, firms using big data analytics are significantly more likely to innovate.  

For Latin America and the Caribbean (LAC), these developments present both opportunities and challenges. In a region where innovation systems are often constrained by low researcher density, limited resources, and weak institutional coordination, AI may help leapfrog traditional innovation barriers. Scaling up the physical capital needed for AI can be achieved more quickly than expanding the pool of highly trained researchers, and this could be an advantage for LAC countries. Despite this potential, research on AI development and adoption in emerging economies remains limited, as research public policies to enhance AI creation and use, thereby accelerating its development.  

Within the international LALICS conference held in Rio de Janeiro in September 2025, the Inter-American Development Bank (IDB) and the UNESCO Chair at the United Nations University UNU-MERIT promoted a panel debate on "Innovation engine: Artificial Intelligence as a catalyst for innovation". The target of the panel was to discuss a research agenda with innovation policy practitioners and academics that reflects the realities and constraints of LAC countries, and that could inform the design and implementation of policies. What do we need to understand to design innovation policies suited to an era of AI-driven innovation? Which institutional, financial, or data-related bottlenecks should be prioritised for research?  We invited a group of distinguished experts from academia as well as innovation policy practitioners, which included Gabriel Yoguel (UNGS, Argentina), Fernando Santiago (UNIDO), Selva Olmedo (UNA, Paraguay), Jocelyn Olivari (CORFO, Chile), and Caetano Penna (Centro de Gestão e Estudos Estratégicos - CGEE, Brazil). Fernando Vargas of the IDB moderated the panel.  

The discussion was grounded in both emerging empirical evidence and forward-looking policy needs, to help to build a regional research agenda that enables LAC countries to fully leverage the potential of AI-enabled innovation.

For example, Brazil launched a comprehensive strategy to strengthen both the production and use of AI, reaching the highest level of related publications in LAC and establishing 144 laboratories with AI developments. However, Caetano Penna of CGEE raised a strategic question that shapes Brazil's approach: recognising the practical constraints of competing for global leadership in AI production, should countries like Brazil prioritise building stronger capabilities in AI adoption and dissemination across sectors? He emphasised that this choice has significant implications for where countries should direct resources and what types of benefits they can realistically expect from their AI strategies – a question of particular relevance for Latin American countries with limited budgets navigating similar strategic decisions.

Chile is also supporting innovation in AI, and CORFO (the country's industrial and innovation development agency) has recorded that the share of projects receiving a subsidy and utilising AI rose from less than 3 per cent in 2010 to 25 per cent in 2024. Interestingly, Jocelyn Olivari reminded the panel that companies have been evolving from basic to more sophisticated uses of AI tools, such as generative AI combined with computer vision, robotics, and AI-based optimisation models. These tools are impacting different industries, with the mining sector being the largest adopter of AI technologies for efficiency and optimisation purposes. Dealing with tail management and the prediction and prevention of potential breakdowns in the extraction and refining processes in real-time are examples of how this sector is utilising AI to address the challenge of low productivity levels.

However, the adoption of AI technologies faces significant structural obstacles. Paraguay, which invests only 0.14 per cent of its GDP in R&D, continues to be constrained by limited technical capabilities, weak digital infrastructure, and insufficient coordination within its innovation system. As noted by Selva Olmedo from the National University of Paraguay (UNA), recent government initiatives, including a new Digitalization Plan (Plan TIC 2022-2030), the National Cybersecurity Strategy 2025-2028, and a draft Law on Artificial Intelligence, aim to strengthen digital governance and promote collaboration between the public and private sectors to support the responsible adoption of AI.

Is AI a possible catalyst for more innovation? Is it a display of a radical transformation or rather the evolution of a paradigm? Gabriel Yoguel, of the Universidad Nacional de General Sarmiento, Argentina, argued that the development of AI requires spaces of interaction among the different actors, thereby including research and education, industry, services, social and government organisations. Such interactions are not often granted. The proper institutional framework is often too weak to allow the process of socio-technical co-creation between humans and artefacts. Many conditions for AI to be a catalyst for innovation are still lacking in many LAC countries.  

Substantial barriers to innovation persist, including the absence of an innovation culture in many firms that should take the process of creative destruction seriously. The challenge is to change the mindset at the corporate level and understand that innovation should be part of the competitive strategy of firms of all sizes and sectors, and that it has a direct impact both in business opportunities and societal development, argued Jocelyn Olivari of CORFO. Infrastructural and organisational conditions pose additional barriers, for example, the lack of interoperability of data among different organisations within Chile's government sector.

Can multilateral organisations play a role in this context? Fernando Santiago, of the United Nations Industrial Development Organisation (UNIDO), argued that AI adoption (like past technologies) does not proceed in leaps, but rather follows a gradual trajectory. In this regard, UNIDO currently supports the adoption of digitalisation agendas within industrial policies in emerging countries. It is essential to understand the necessary conditions for the dissemination and adoption of AI, including its integration with other advanced digital production technologies in manufacturing firms. The likely unbalanced distribution of capabilities for digitalisation across firms, countries, and regions is of concern. Latin American countries are still in the early stages of developing strategies to capitalise on the opportunities associated with AI in terms of efficiency and productivity gains.  

The panel shared the view that public policies must always be evidence-based to be effective and contribute to achieving the goals that societies set forth. Research, both theoretical and applied, can play a crucial role in this regard, enhancing governments' ability to act effectively. Given the speed and the likely spread of AI across economies and societies, what are the pressing questions that Governments face and that research should address?  

The discussion surfaced two distinct, but interconnected, sets of questions: (i) some concerning AI's broader developmental and institutional implications, and (ii) others centred on AI's role in innovation and firm transformation.

Among the first,  

• AI as a development enabler: How, and under what conditions, can AI contribute to a country's socioeconomic development? What mechanisms will shape the distribution of its benefits and opportunities across sectors and social groups?
• Socio-technical co-creation: The creation and adoption of AI technologies involve co-creation between humans and artefacts. Which models of socio-technical interaction are most effective? What skills and capabilities are most needed to sustain these processes?
• Governance and coordination: Policies to foster AI innovation (like all public policies, though perhaps even more so) require the coordination of multiple actors and the alignment of their interests. What forms of governance best enable coherent and collaborative policy implementation?

And the second,

• Fostering innovation investment: How can governments encourage firms to invest more in innovation, particularly in AI-related technologies, robotics, and data sciences? How can a culture of AI-enabled innovation be created, nurtured, and disseminated?
• Firm-level adoption and transformation: AI adoption entails big organisational change. What are its effects on firm performance, particularly among SMEs? How can policy help overcome adoption barriers and support sustainable transformation?

The discussion revealed a shared sense of urgency. Latin American countries cannot afford to remain passive observers of a technological transformation that is reshaping the global knowledge frontier. As AI accelerates scientific discovery and innovation, it is also raising the opportunity cost of maintaining lagging innovation systems. The policy practitioners and researchers who joined this conversation were not seeking universal recipes, but articulating questions rooted in the realities, constraints, and development goals of their countries. Turning those questions into a collaborative research agenda should now be a priority for scholars in this field, helping to build the evidence that can truly guide policy in an era of AI-driven innovation.