New research from StackGen, analyzing nearly 178,000 public technology incidents, reveals that AI-related outages now make up more than one in 10 reported disruptions, a rate six times higher than in 2023. These incidents often involve AI systems autonomously deleting data, databases, or live systems using valid credentials, which means traditional monitoring tools failed to detect the issues until after the damage was already done. As AI becomes more deeply integrated into business operations, the risk of such outages is growing significantly. AI is now embedded in a wide range of business functions, from claims processing and customer support to fraud detection and supply chain planning. This integration introduces hidden risks, as AI systems can create dependencies that are difficult to track or understand. These risks extend beyond just AI-powered workflows, as AI also influences broader enterprise operations and decision-making processes. As a result, companies are facing a growing challenge in managing the potential consequences of AI failures. The use of AI also increases the risk of cyber threats. Attackers can now use AI to create deepfakes, scale phishing campaigns, and develop more sophisticated cyberattacks. Additionally, AI systems can change their behavior over time, leading to unpredictable outcomes. This "drift" in AI behavior can make it difficult to understand or predict system performance. Meanwhile, employees may use unapproved AI tools—often called "shadow AI"—in critical workflows, introducing additional risks that are hard to monitor or control. Recovery from AI-related disruptions can be particularly complex. If an AI-enabled workflow fails or becomes unavailable, organizations may struggle to determine the full impact or implement a manual fallback. This lack of preparedness can leave businesses vulnerable during critical moments of disruption. To address these challenges, organizations must understand the full scope of AI dependencies, the potential consequences of their failure, and the steps needed to ensure resilience. As AI becomes more central to business operations, companies must decide which processes can tolerate "probably right" answers and which require absolute certainty. For example, AI-generated recommendations for decisions may be acceptable if they are "probably right," but processes involving financial transactions or critical operations must be predictable and repeatable. Organizations must also evaluate AI risk through four key questions: what is impacted, what happens next, what is the financial exposure, and what should be prioritized in terms of controls and oversight. Explainability—understanding how and why AI models reach their conclusions—is becoming a crucial factor in AI adoption. If a vendor cannot clearly explain its AI model's behavior, it can expose an enterprise to legal and compliance risks. Recent lawsuits against companies like Cigna and Workday show that businesses can be held accountable for AI-related issues, even if they did not develop the technology themselves. Mapping AI dependencies is essential to prevent single points of failure. Just as businesses monitor their supplier relationships, they must also track how AI models, data sources, and third-party applications influence their operations. A disruption in one area, such as a cloud service or model provider, can ripple across multiple business functions. Understanding these relationships helps organizations better assess the potential impact of AI failures. Finally, enterprises must treat AI agents as critical assets, keeping track of their functions, ownership, and dependencies. Failing to manage these agents can lead to compliance issues, uncontrolled costs, and blind spots in incident response. By cataloging AI agents and understanding their role in business processes, organizations can build a more resilient and secure AI environment. As AI adoption continues to grow, enterprise resilience will be key to managing the risks and ensuring business continuity.