Amazon’s AI boom sparks internal chaos with duplicate tools and data

Amazon's AI surge is generating a new problem: An increasing amount of internal tools and repeated data.

Some teams are quickly developing their own AI-driven applications to streamline workflows and manage data.

While this wave of innovation is exciting, it also causes issues, such as software and data duplication, according to an internal document that Business Insider has reviewed.

"AI is exacerbating our issue of tool duplication," the document noted. "Tools are being duplicated more rapidly, with less cleanup occurring."

This trend reflects a larger movement across businesses in the United States. Generative AI is causing what is being termed "AI sprawl," a surge of AI tools and independent agents that can overwhelm firms' centralised oversight and security measures.

As more staff members independently create tools, sometimes within minutes using AI helpers, businesses may lose track of the systems in use, the locations of sensitive data, and how much duplicate software they possess.

It's a well-known narrative with a new angle. Two decades ago, with the rise of cloud computing, employees would open Amazon Web Services accounts without authorisation.

Subsequently, "SaaS sprawl" distributed cloud applications throughout companies with minimal initial oversight.

In both scenarios, businesses later reacted by curbing excess, implementing oversight, and crafting new methodologies to officially harness these potent innovations. Generative AI is undergoing a similar trajectory, albeit at a much faster pace.

The consequences of this disordered, innovative AI upswing could be significantly noticeable at Amazon.

The corporation has integrated AI across its operations, with CEO Andy Jassy motivating employees to integrate this technology or risk falling behind.

Amazon has long faced the issue of teams simultaneously developing similar tools, according to the document accessed by Business Insider.

These overlaps were costly but manageable. Software development used to require substantial time and engineering resources, and the burden of maintaining it led to some redundant tools being eventually phased out.

As per the document, AI "significantly reduces the barriers to crafting new tools," allowing teams to prototype and release software much more swiftly.

Instead of searching for existing solutions, engineers can create their own in a shortened timeframe and sustain them at a reduced cost.

This has resulted in a rise of overlapping systems, with reduced impetus to unify them. "AI is now worsening this issue from both ends," the document stated.

The problem extends beyond just tool overload. It's also creating new challenges regarding data management and storage.

Numerous AI systems at Amazon use internal data to create new formats, such as knowledge bases and summaries.

These new iterations are often stored independently of the original data, forming new duplicates according to the document.

If the original data is eventually deleted or access is restricted, those produced versions don't always update.