July 6, 2026, (Inside AI) — A new study reveals that artificial intelligence significantly bolsters firm resilience during natural disasters, but its protective shield falters against human-caused crises. The research, led by Professor Wu Jing of CUHK Business School, analyzed 3,137 U.S. firms across eight sectors, finding that companies with at least 2.4% of job postings requiring AI skills recover faster and suffer milder stock losses.
The paper, titled Artificial intelligence and firm resilience: Empirical evidence from natural disaster shocks, was co-authored with Michael Zhang, Han Miaozhe of Hong Kong University of Science and Technology, and Shen Hongchuan of the University of Macau. It arrives as businesses face a polycrisis of climate disruptions and operational shocks, challenging the narrative that AI is a universal risk buffer.
“AI generates significant resilience for firms facing natural disaster shocks primarily by optimising supply chains and production inputs,” Professor Wu said. The effect peaks during the height of a disaster, with AI-empowered roles in cognitive tasks, decision-making, and supply chain coordination proving most effective.
But the armor cracks when crises stem from human actions. “Such resilience may not persist in the face of human-induced shocks, such as cyberattacks, labour strikes, or industrial accidents,” Wu noted. In those scenarios, AI serves only as a risk detector, not a mitigator, because reputational and contractual damage can’t be algorithmically reversed.
Read: AI Undermines Leaders' Judgment
The study distinguishes AI from general information technology. While IT improves efficiency through coordination and monitoring, AI’s predictive power is what drives resilience. The researchers measured non-AI tech investments via job postings with skills like robotics or cloud computing, and found IT boosts daily operations but doesn’t match AI’s crisis-shielding role.
“AI serves as complementary support for tangible operations and works more efficiently if it targets physical assets, such as factories,” Wu explained. “When AI is used on financial assets, intellectual property, or market access, its effects are limited.” This insight challenges the blanket application of AI across all business functions.
Financially constrained firms gain the most relative benefit from AI during crises, yet their productivity still lags behind wealthier peers in normal times due to weaker organizational infrastructure. Wu advises these companies to treat AI as an insurance premium, not a profit engine, and to first build the IT backbone and training procedures needed to deploy it effectively.
The Resilience Threshold and Strategic Deployment
The 2.4% AI-hiring threshold marks a clear line for full recovery. Firms below it see prolonged losses. The finding echoes research from the National Bureau of Economic Research showing that AI adoption widens productivity gaps between leaders and laggards, especially in capital-intensive industries.
“Roles like supply chain coordinators should be prioritised to be empowered with AI skills for better predicting materials arrivals and planning alternative routes,” Wu said. He also urged equipping strategic decision-makers and production managers with AI tools for quicker resource allocation during shocks.
Competing viewpoints caution against over-indexing on AI for resilience. A 2025 McKinsey report found that resilience depends more on supply chain redundancy and workforce agility than on any single technology. Yet the CUHK study’s empirical rigor—using stock returns as a real-time performance proxy—lends weight to its sector-specific prescriptions.
Read: Why Frontier AI Still Hallucinates
The research also raises questions about AI’s role in compound crises. If a hurricane triggers a chemical spill, the interplay of natural and human-caused shocks could blunt AI’s effectiveness. The study’s authors acknowledge this limitation, noting that their data predates the rise of generative AI tools, which may alter the resilience calculus.
Beyond the Hype: What Firms Should Actually Do
For cash-strapped firms, the message is clear: don’t just buy AI tools—redesign workflows. “These firms should view AI investment as an insurance premium for resilience instead of an immediate profit engine, ensuring they have the IT backbone to support it,” Wu added. He recommends deploying AI first in vendor monitoring or financial planning to generate savings that fund further resilience measures.
The study’s focus on U.S. firms limits its global applicability. Supply chain dynamics in regions with weaker digital infrastructure—such as parts of Southeast Asia or Africa—may not see the same AI-driven resilience. A 2024 World Bank report found that AI adoption in developing economies is often hampered by data scarcity and regulatory gaps.
Looking ahead, the researchers plan to examine how generative AI and large language models affect resilience. Early evidence suggests these tools could enhance decision-making speed but also introduce new vulnerabilities, such as prompt injection attacks or hallucinated risk assessments.
“If crisis is the new norm, infusing AI into firm productions is no longer a luxury,” Wu concluded. The study serves as both a playbook and a warning: AI can weather some storms, but not all.
