The Countries Best Prepared for AI Are Pulling Ahead in Eco-Innovation

Artificial intelligence can help firms design cleaner products, use resources more efficiently and accelerate green innovation, but the gains depend heavily on the digital environment around them, warns a new study Digital Economies and Artificial Intelligence: Unlocking Eco-Innovation in Asian Firms by Marwan Mansour, Ismail Younes Yamin, Abdulrahman Alomair and Mohammed Alomair, published in the journal Computation.

Using 44,512 firm-year observations from 5,564 listed companies across 15 Asian economies between 2017 and 2024, the researchers find that stronger AI capability is associated with higher eco-innovation. The effect becomes more powerful in countries with better digital infrastructure, connectivity, skills, business digitalization and public digital systems.

The study challenges a narrow policy focus on AI adoption. Algorithms and automation do not generate environmental innovation in isolation. Firms need organizational capacity, while countries need digital ecosystems capable of turning technological investment into cleaner products, processes and business models. The findings also carry a warning for developing economies: without investment in infrastructure and skills, AI could widen the innovation gap between large and small firms and between digitally advanced and lagging markets.

AI Capability Is Becoming a Green Innovation Asset

Eco-innovation includes products, production methods and management practices that improve environmental performance while preserving commercial competitiveness. AI can support these activities by processing operational and environmental data, identifying inefficiencies and improving decisions. Machine-learning tools can help firms detect patterns in energy use, optimize material inputs, forecast equipment problems and assess the environmental consequences of design choices.

The study measures AI capability through textual analysis of annual and sustainability reports. It tracks terms related to artificial intelligence, machine learning, automation, robotics and predictive analytics, adjusting for report length. Eco-innovation is measured through a Refinitiv score covering roughly 20 indicators related to green products, emissions reduction, resource efficiency and sustainable technology.

In the full fixed-effects model, AI capability retains a positive and statistically significant coefficient of 0.121. The relationship remains positive after the authors use System Generalized Method of Moments estimation, propensity-score matching and a Heckman selection model to address reverse causality, selection effects and the possibility that innovative firms are simply more likely to invest in AI.

The finding positions AI as more than an efficiency tool. When connected to environmental strategy, it can help firms discover new products, improve production systems and incorporate sustainability information into commercial decisions. The same data infrastructure that supports productivity can also reduce waste, emissions and resource use.

However, the study does not show that every AI investment delivers an environmental benefit. A company that purchases software or highlights AI in its reports may achieve little if it lacks reliable data, technical staff and clear sustainability objectives.

The Real Multiplier Sits Outside the Firm

The paper focuses on the national digital economy as a condition shaping firm-level results. The researchers construct an index covering digital skills, infrastructure, business digitalization and public digital services using data from the World Bank, International Telecommunication Union, UN E-Government Development Index and Network Readiness Index.

The interaction between AI capability and digital economy development is positive and statistically significant. The interaction chart on page 18 shows that eco-innovation rises more sharply with AI capability in highly developed digital environments than in weaker ones.

Firms in digitally advanced economies have more reliable connectivity, richer data resources, stronger knowledge networks and larger pools of skilled workers. They are also more likely to benefit from public systems and regulations that support technology adoption and data use. In weaker ecosystems, firms may struggle to recruit specialists, obtain usable data or connect new systems with suppliers and public agencies.

This shifts responsibility beyond corporate management. Governments influence AI's environmental payoff through broadband access, digital education, data governance, research systems and public-sector capacity. National digital policy is therefore also industrial policy and, increasingly, climate policy.

For Asian economies pursuing digitalization and decarbonization together, sequencing matters. Encouraging AI adoption before addressing infrastructure and skills gaps may produce fragmented projects with limited environmental value. Building shared digital foundations can make private investment more productive and allow more firms to participate.

The implication is especially important for the Global South, where public resources are constrained. Digital infrastructure should not be treated only as a communications investment; it can also support industrial upgrading and progress toward SDG 9 on industry and innovation, SDG 12 on responsible production and SDG 13 on climate action.

AI's Green Dividend Risks Becoming a Large-Firm Privilege

The estimated effect of AI capability on eco-innovation is 0.041 for large companies, compared with 0.019 for smaller companies. Both relationships are positive, but larger firms appear better able to convert AI capacity into environmental innovation.

Large companies generally have more finance, stronger research departments, better data systems and greater access to specialized talent. They can absorb implementation costs and integrate AI across product development, manufacturing, supply chains and environmental management.

Smaller companies often lack structured data, dedicated technology teams and the capital needed to experiment with uncertain returns. The study's listed-firm sample also means its "smaller" companies are not necessarily representative of the micro and informal businesses that dominate employment in many Asian economies. The true capability gap may be wider.

This creates a policy dilemma. Tax incentives or subsidies for individual AI purchases may flow disproportionately to companies already equipped to benefit. Large firms could accelerate green innovation while smaller suppliers and competitors fall further behind.

A more inclusive approach would provide shared services. Industrial parks, business associations and public innovation agencies could offer common data platforms, cloud resources, environmental analytics and technical advice. Development banks could combine concessional finance with implementation support rather than fund hardware alone.

Multilateral institutions can also promote interoperable standards, regional skills programmes and cross-border research networks. Because Asian production systems are connected through supply chains, helping smaller suppliers meet digital and environmental requirements could spread gains beyond leading corporations.

The Evidence Is Strong, but the Green Balance Sheet Is Incomplete

The study's multi-country scope is a major strength. Much previous research has focused on China, while this analysis covers firms across 15 Asian economies with different institutional and technological conditions. Its large panel and multiple robustness tests also strengthen confidence in the central relationship.

Important limitations remain. AI capability is inferred from corporate disclosures, which may capture communication strategy as well as actual deployment. Companies skilled at reporting may appear more AI-ready than firms using the technology operationally but discussing it less.

The sample excludes most private firms, start-ups and small businesses. The paper also contains an internal inconsistency: its abstract describes the panel as unbalanced, while the methodology calls it balanced. That point should be clarified.

Most importantly, the study measures eco-innovation rather than AI's net environmental impact. Data centres, advanced chips and digital infrastructure consume electricity, water and materials. AI may help a firm create a more efficient product while increasing computing demand elsewhere.

Future research should distinguish among generative AI, machine learning, computer vision, robotics and predictive analytics because their environmental uses and energy demands differ. It should also combine disclosure measures with patents, capital spending, operational deployment and verified environmental outcomes.

Asia's AI-powered green transition will be decided less by how many firms announce technology projects than by how effectively countries convert digital capacity into broad-based sustainable innovation. Without that foundation, AI may deepen existing divides. With it, the technology could become a practical engine of cleaner industrial growth.