Amazon Admits Heavy AI Use Is ‘Disrupting’ Its Core Business
Tech giants are learning the hard way that forcing AI into every workflow can trigger a massive backfire. Amazon recently summoned a large group of engineers to address a series of damaging outages plaguing its online retail business. The company admitted that its aggressive push into generative AI is now “wreaking havoc” on its core operations.
The internal alarm was sounded by Dave Treadwell, a senior vice president at Amazon’s eCommerce Services, who told staff that site availability has “not been good recently”. This admission follows a nearly six-hour outage last week that left customers completely unable to place orders. The culprit was a botched software code deployment linked to AI-assisted changes.
While Amazon remains committed to the technology, it is now grappling with a “trend of incidents” that have a high “blast radius”. The company is struggling to balance its speed of innovation with the reality of unstable systems. The result is a chaotic environment where automated tools are moving faster than the humans meant to supervise them.
The High Blast Radius
The chaos isn’t limited to the shopping site; it has also hit Amazon Web Services (AWS), the company’s massive cloud division. Internal reports revealed two separate outages caused by engineers allowing in-house AI coding tools to make disastrous changes. In one particularly alarming case, the AI tool actually deleted and then recreated an entire coding environment.
Amazon described these failures as the result of “novel GenAI usage” where best practices and safeguards simply do not exist yet. Because generative AI models frequently allow errors to slip through or struggle with complex instructions, they can take destructive actions that a human user never intended. The technology is proving to be as volatile as it is powerful.
Despite these blunders, Amazon is sticking to its guns and refuses to back away from AI deployment. Instead, the company is framing the issue as a failure of “user access control” rather than an inherent problem with AI autonomy. The solution, according to leadership, is more oversight—even as the number of people available to provide that oversight shrinks.
Fewer Humans, More Problems
The timing of this AI crisis creates a sharp contradiction within Amazon’s corporate strategy. Management is leaning on programmers to use AI tools for 80% of coding tasks every week. However, this renewed demand for strict human supervision comes at the exact same moment the company is aggressively slashing its workforce.
Amazon has already fired hundreds of workers from its cloud computing division and is currently targeting a massive layoff of 30,000 corporate employees. This creates a mathematical nightmare: the company wants more AI coding and more human oversight, but it is doing so with significantly fewer humans. The remaining staff must now police more code in less time.
Junior and mid-level engineers are now required to have senior engineers sign off on every single AI-assisted change. This adds a new layer of bureaucracy to a system that was supposedly designed to increase efficiency. The “blast radius” of AI errors is growing, but the safety net of human expertise is being pulled away.
A Risky Path Forward
Amazon’s current predicament serves as a warning for the entire tech industry. Rapidly deploying tools before established guardrails are in place can jeopardize the very infrastructure that keeps a business alive. For now, the e-commerce giant is betting that stronger “guardrails” will be enough to prevent another site-wide collapse.
The core business of Amazon relies on 100% availability for its shoppers and cloud clients. Every hour of downtime translates to millions of dollars in lost revenue and a massive hit to consumer trust. If the trend of AI-related incidents continues, the cost of “efficiency” may eventually outweigh the benefits of the technology.
Whether Amazon can successfully navigate this transition with a skeleton crew remains to be seen. For a company that prides itself on precision, the current state of “havoc” is a rare and public stumble. The future of the shopping site may depend on whether the AI or the humans end up in the driver’s seat.