Amazon has an AI problem, and it's not what you'd expect. An internal document obtained by Business Insider reveals that generative AI is causing tool duplication to spiral across the company's retail division. "AI is making our tool duplication problem worse," the document states plainly. "More duplication is being created faster, and less of it being cleaned up." The math is simple: copy-and-paste AI and later stages of adoption make it so easy to spin up new tools that engineers would rather build their own than find existing ones.
The risks go beyond bloat. Amazon's AI systems ingest internal data and convert it into knowledge bases, summaries, and other outputs. Those derived artifacts get stored separately from the originals. When source data gets deleted or restricted, the AI-generated copies stick around. In one documented case, a tool called Spec Studio kept displaying software details that had been made private in Amazon's code repository. Debo Dutta, chief AI officer at Nutanix, told Business Insider that ungoverned "shadow AI" deployments can lead to data exposure and regulatory trouble.
Amazon's proposed fix is to fight AI with AI, using machine learning to identify duplicate tools and flag risks early. The real tension runs deeper. Amazon's famous "two-pizza team" model gives small groups autonomy to move fast. Speed comes at the expense of organizational change required to scale. CEO Andy Jassy has been pushing employees to adopt AI aggressively, warning they'll fall behind if they don't. Move fast. The document shows what happens next.
Other big tech companies took different paths. Meta consolidated its fragmented AI efforts into a single product group in early 2023, building shared models like Llama to avoid redundant work. Google merged Brain and DeepMind. Microsoft relies on centralized Azure AI governance. Amazon bet on autonomy. That bet produced a mess growing faster than anyone can clean up.