Amazon is grappling with a significant internal challenge known as AI sprawl, a phenomenon where the rapid adoption of generative AI tools has led to excessive software duplication and data fragmentation. According to a confidential document obtained by Business Insider, employees across the company are increasingly building their own AI-powered applications to automate workflows, creating a complex web of redundant systems that outpace centralized oversight. The document highlights that while Amazon has historically faced issues with parallel tool development, the barrier to entry for creating software has been dramatically lowered by artificial intelligence. Engineers can now prototype and deploy applications in a fraction of the time it previously required, often without needing to check for existing solutions. Consequently, the pressure to consolidate redundant tools has diminished, leading to a surge in overlapping systems that are difficult to manage. This trend mirrors previous technological shifts, such as the rise of cloud computing and SaaS products, where unauthorized accounts and software eventually created visibility gaps for IT departments. However, the speed at which generative AI enables tool creation is accelerating this cycle. The document warns that organizations are losing visibility into which systems are active, where sensitive data resides, and the extent of their software redundancy. Beyond tool proliferation, Amazon faces specific data integrity risks. Many AI systems ingest internal data, transform it, and store the output in separate formats, effectively creating persistent copies of information. A critical issue identified is that these derived artifacts often do not update when the original source data is deleted or access permissions are restricted. For example, a tool called Spec Studio reportedly continued displaying software details that had been made private in Amazon's code repository. This persistence of derived data poses security and compliance risks, as sensitive information may remain accessible long after it should have been removed. Amazon's leadership has acknowledged the tension between the company's culture of autonomy and the need for control. CEO Andy Jassy has urged employees to adopt AI to remain competitive, yet this encouragement has inadvertently fueled the sprawl. The company's long-standing "two-pizza team" model, which empowers small, independent groups to make rapid decisions, is now making coordination more difficult. While the document reflects the perspective of a specific team evaluating AI tools in the retail business, an Amazon spokesperson noted that it does not necessarily represent the experience of the entire workforce. To address these issues, Amazon is exploring the use of AI itself as a solution. The company is investigating methods to leverage AI to identify duplicate tools, flag potential risks, and encourage teams to consolidate their efforts before overlaps become unmanageable. Experts outside the company warn that without proper governance, the unchecked spread of AI tools, or shadow AI, could lead to significant data exposure and system disruptions. As Amazon navigates this new landscape, balancing the drive for rapid innovation with necessary centralized security and efficiency remains a critical challenge.