Amazon's aggressive internal push to adopt artificial intelligence is creating a significant operational challenge: a surge of duplicated software tools and fragmented data copies that are eroding central oversight and control. According to an internal document obtained by Business Insider, the ease of building AI applications is exacerbating long-standing issues of tool duplication within the tech giant's vast retail division.
The problem, termed "AI sprawl," reflects a broader corporate trend where generative AI tools are proliferating faster than governance frameworks can manage. At Amazon, this is leading to increased costs, security vulnerabilities, and persistent copies of sensitive data that may not update when source information changes.
Lowered Barriers Worsen Existing Problems
For years, Amazon has managed the cost and complexity of teams building similar tools in parallel. However, AI "dramatically lowers the barrier to building new tools," states the confidential February document from a team evaluating AI tools in Amazon's retail business. This allows engineers to prototype and deploy solutions rapidly, with far lower maintenance costs, reducing the incentive to find and use existing company-wide systems.
"AI is now making this problem worse from both directions," the document concluded, noting that more duplication is being created faster and less is being cleaned up. The company's celebrated "two-pizza team" culture of autonomous, fast-moving units may be intensifying the issue.
Data Duplication Introduces New Security Risks
The sprawl extends beyond software to critical data management. Many internal AI systems ingest company data, transform it into summaries or knowledge bases, and store these outputs separately. This creates derived copies that may not reflect changes to the original source, such as updated access permissions or deletions.
The document cited one internal case where a system called Spec Studio continued displaying software details that had been made private in Amazon's internal code repository. "Any system that ingests data, transforms it through AI, and stores the output separately faces the same problem," the analysis warned.
Corporate America's Familiar Pattern with a New Twist
The situation mirrors previous technological shifts, such as the emergence of cloud computing and SaaS software, which initially led to similar "sprawl" before companies established governance. Debo Dutta, chief AI officer at Nutanix, told Business Insider that this rapid spread creates "shadow AI," where unauthorized applications can lead to data exposure and regulatory violations.
"If not governed properly, this can all lead to data and system disruption," Dutta said. The key challenge for Amazon and others is balancing the speed and innovation of AI adoption with necessary coordination and control.
Exploring AI-Driven Solutions
Amazon's potential solution to the problem may involve using more AI. The document indicates the company is exploring ways to leverage AI to identify duplicate tools, flag risks, and encourage consolidation earlier in the development process. This approach aims to maintain the pace of innovation while mitigating the downsides of uncontrolled tool creation.
In an email to Business Insider, Amazon spokesperson Montana MacLachlan stated that the document reflects the perspective of a single team and that it is "inaccurate" to use it to characterize the experience of the company's broader workforce. The company encourages internal dialogue on such challenges as part of its problem-solving culture.