Global Economic Impacts of AI

Our 2025 winter Buffett Symposium on AI and Geopolitics convened leading strategists, researchers, and policymakers to discuss the transformative opportunities and profound challenges that AI poses in geopolitics. The event was co-organized by the Buffett Institute for Global Affairs, ⁠Northwestern Security & AI Lab (NSAIL)⁠, and ⁠Insight Centre at University College Cork⁠.

The daylong program's penultimate panel discussion focused on global economic impacts of AI. The economic disparities in AI adoption across regions and industries are influenced by factors such as regulatory environments, infrastructure readiness, and cultural attitudes toward risk. The panel discussed the barriers to entry of developing and deploying use-case specific enterprise AI systems, including operational agility and compliance with regulatory environments, while acknowledging the low-hanging fruit of enhancing white-collar workforce productivity, optimizing operations, and automating customer service.

Panelists included:

• David Bray, Distinguished Fellow and Chair of the Loomis Accelerator with the Alfred Lee Loomis Innovation Council at the non-partisan Henry L. Stimson Center, and former Chief Information Security Officer at the US Federal Communications Commission

• Johan Harvard, Global AI Advisory Lead at the Tony Blair Institute for Global Change in London

• Sandeep Mehta, Advisory Board Member of the Ethical AI Governance Group, and former Chief Technology Officer at the Hartford Financial Services Group

• Moderator Daniel W. Linna Jr., Senior Lecturer and Director of Law and Technology Initiatives at Northwestern University's Pritzker School of Law

Key Takeaways

• The sector-specific variability in the return on investment (ROI) of AI reflects the substantial investment in infrastructure, data standardization, and workforce training to make the systems effective. Harvard explained that the ROI calculus is improving AI’s economic incentives for many stakeholders, making it worthwhile for certain sectors, but the upfront effort remains a prohibitive barrier. Rather than expecting quick wins from an out-of-the-box solution, achieving meaningful ROI with AI requires a strategic, long-term approach.

• While AI offers real productivity gains, the expectations surrounding its transformative power may be over-hyped. AI applications provide tangible, incremental improvements to existing systems and workflows, Mehta noted, but they are not so-called "killer apps.” Mehta reported that, in the finance sector, bullish projections estimate 30 percent productivity gains, but the actual gains are measured at five or six percent.

• The focus on generative AI has diverted attention and resources from other promising approaches. Unlike deep learning, which requires extensive training on vast datasets, active inference is modeled after how humans learn and make predictions with limited information. Bray noted that active inference may offer more privacy-preserving, energy-efficient, and data-efficient solutions, but has been overshadowed by the sunk costs and focus on generative AI, limiting its exploration and adoption.

Read the symposium synthesis report >>