Artificial intelligence (AI) is best understood not as a single technology but as a set of processes that make existing activities smarter, faster, and more efficient. These processes combine data analytics, enabling technologies, software, and applications. Existing uses include self-driving cars, speech recognition and translation, supply chain optimization, and medical diagnostics.

There are differences between narrow AI and general AI. Narrow AI – today’s reality – uses machine learning to perform specific tasks, while general AI, which would replicate broad human intelligence, remains speculative and raises longer-term existential concerns. Narrow AI relies on machine learning models trained on large volumes of data to generate increasingly accurate predictions.

Several technical developments underpin modern AI, particularly deep neural networks (DNNs). These systems stack multiple layers of data processing and can integrate images, text, audio, and video, creating general-purpose machine learning. Equally important are complementary tools such as reinforcement learning, graphics processing units (GPUs), and techniques for validating and training models at scale. Together, these advances explain why AI can now operate on the ‘chaotic data’ humans routinely process.

The economic impact of AI

AI could add trillions of dollars to global output. McKinsey estimates that AI could increase global GDP by around 16%, or US$13 trillion, by 2030, while other studies suggest AI could double annual economic growth in advanced economies by 2035.

The largest gains are expected to come from productivity growth, but productivity effects take time. Like earlier general-purpose technologies, AI requires complementary investments in capital, skills, and organizational change before aggregate benefits materialize. This lag helps explain why productivity growth remains low despite rapid technological advances.

AI will also change the composition of growth. It is likely to accelerate the shift toward a service-driven economy and intensify demand for high-skilled labor. In US manufacturing, for example, the share of high-skilled jobs rose from 6% in 1930 to 30% in 2010, while low-skilled jobs declined sharply. AI’s application across sectors, from finance and transportation to health care and education, underscores its economy-wide reach.

The impact of AI

On global value chains (GVCs)

AI affects trade through both macroeconomic and firm-level channels. By boosting productivity and competitiveness, it creates new opportunities for cross-border exchange. As services become more valuable, their share in international trade is likely to grow further.

AI is already reshaping trade logistics and global value chains (GVCs). Firms use AI to manage supply chain risks, forecast demand, improve warehouse operations, and enable smart manufacturing. The German concept of Industry 4.0 – integrating sensors, the internet of things, and cyber-physical systems – illustrates how AI can enable predictive maintenance, mass customization, and tighter coordination across borders.

However, AI may also encourage reshoring. Automation and technologies such as 3D printing reduce reliance on low-cost labor and long supply chains, potentially accelerating “premature industrialization” in developing economies.

On e-commerce

Digital platforms show AI’s immediate trade effects. On eBay, 97% of US small businesses export, compared with only 4% of offline peers. AI-powered machine translation has amplified these gains: eBay’s translation tools increased US exports to Spanish-speaking Latin America by 17.5%, with export values rising 13.1%. This is equivalent to reducing geographic distance between countries by more than 35% – a striking illustration of AI’s trade-enhancing power.

On trade negotiations

AI can also improve trade policymaking itself. Governments can use AI to model negotiation outcomes, assess multi-country scenarios, and predict third-country responses. Brazil’s Intelligent Tech & Trade Initiative demonstrates how AI can become a strategic asset in trade diplomacy.

AI’s potential economic and social costs

AI’s benefits are not automatic. Transition costs may include job displacement, wage polarization, and increased inequality. While only about 9% of jobs are at risk of full automation, low-skilled workers are disproportionately impacted. The OECD estimates that 44% of US workers without a high school degree hold highly automatable jobs, compared with just 1% of workers with a bachelor’s degree or higher.

Developing countries face mixed outcomes. Early AI adoption in advanced economies could erode manufacturing-based comparative advantage elsewhere, while gaps in skills, cloud infrastructure, and data access risk widening the global digital divide.

Policies to regulate and facilitate AI: Seven action areas

Strengthen AI diffusion within and across countries. AI’s benefits depend on broad adoption. Yet productivity gaps between frontier firms and laggards are widening, reflecting uneven diffusion. Policies should promote competition, labor mobility, access to risk capital, and technology transfer to avoid AI gains accruing only to a few firms and countries.
Develop education and skills. AI requires both STEM skills and complementary human capabilities. Education systems must prepare workers to collaborate with AI, while international partnerships can help spread skills development to developing economies.
Establish sound cybersecurity. Cybersecurity is essential for trust in AI. Weak defenses in one country can have global effects. National strategies, such as the US National Institute of Standards and Technology Cybersecurity Framework, are critical, but so is international cooperation to avoid cybersecurity standards becoming disguised trade barriers.
Protect the privacy of personal data. There is no inherent trade-off between AI and strong privacy rules, but poorly designed regulations can restrict data flows. Interoperability between privacy regimes – through mechanisms like the US–EU Privacy Shield – and commitments to free data flows are essential for AI development.
Increase access to government data. Open government data can become a platform that fuels innovation. Domestic data strategies and international frameworks, supported by trade agreements like the US-Mexico-Canada Agreement (USMCA), can reduce the cost of accessing public data globally.
Develop a balanced intellectual property framework. AI training often requires copying data. Flexible copyright exceptions such as fair use in the US support AI innovation, but many countries lack similar provisions. Trade agreements should recognize the need for balance between intellectual property (IP) protection and innovation.
Develop AI ethical principles. Trust in AI requires ethical systems that respect human dignity, transparency, and fairness. Without international coordination, ethics frameworks risk becoming fragmented and discriminatory. Efforts by the Organisation for Economic Co-operation and Development (OECD) and the Group of 20 (G20) could provide a shared baseline, with trade agreements addressing testing standards and compliance.

Access the full white paper