Amazon retail website hit by multiple system failures from outdated AI advice

Amazon faced multiple disruptions on its retail website after engineers reportedly followed inaccurate advice generated by an AI agent that relied on outdated internal documentation. The incidents have triggered internal reviews of AI-assisted development and raised questions about the balance between automation and human oversight in large-scale technology systems.

Amazon experienced a series of major technical disruptions on its retail platform, with four high-severity incidents occurring within a single week.

One of the most serious outages lasted nearly six hours, preventing customers from accessing checkout, viewing account information and seeing product pricing.

The incidents prompted the company to review the causes behind the failures during a regular weekly technology operations meeting involving senior leaders responsible for the company’s ecommerce infrastructure.

AI Advice from Outdated Wiki Linked to Incident

Initial reports suggested that AI tools may have played a role in the outages. According to internal discussions, engineers may have relied on guidance produced by an AI agent that had drawn information from an outdated internal wiki.

Amazon later clarified that no AI-generated code was deployed directly, but acknowledged that an engineer had followed incorrect advice inferred by the AI system.

The company stated that only one of the incidents involved AI tools, and that the meeting reviewing the outages was part of its routine operational review process rather than an emergency gathering.

Internal Documents Show Concerns Around AI Safeguards

Internal communications suggested that the disruptions triggered broader discussions about how generative AI tools are used within engineering workflows.

Senior leadership reportedly acknowledged that best practices and safeguards for generative AI deployment were still evolving. As a result, the company plans to introduce additional oversight and “controlled friction” when AI-assisted changes affect critical retail systems.

The goal is to ensure stronger verification and human supervision when deploying changes to sensitive infrastructure.

Rising AI Investments and Workforce Changes

The incidents come at a time when Amazon is significantly expanding its investment in artificial intelligence infrastructure.

The company is projected to spend around $200 billion on AI-related infrastructure this year, one of the largest technology investments globally.

At the same time, Amazon has been reducing its corporate workforce. The company cut about 14,000 corporate roles in October, followed by another 16,000 layoffs in January, adding to the more than 27,000 jobs eliminated between 2022 and 2023.

Executives have previously suggested that AI-driven productivity improvements could allow the company to operate with fewer employees.

AI Productivity Debate Continues

Amazon’s system failures highlight a broader debate around the real-world impact of AI on workplace productivity.

Some technology leaders argue that AI will significantly reduce the need for human workers. For example, executives at Block, Inc. and Salesforce have publicly linked workforce reductions to AI-driven efficiency gains.

However, recent workplace studies suggest a more complex reality.

Research analysing 164,000 employees found that AI tools often increase the speed, complexity and volume of work, rather than reducing it. Workers reportedly spend more time managing emails, chat systems and digital communication after adopting AI tools.

Gap Between AI Potential and Real Deployment

Further research from Anthropic indicates that the gap between what AI can theoretically automate and what organisations actually automate remains large.

While studies suggest that up to 94% of tasks in fields like software and mathematics could be automated, only about one-third of those tasks are currently handled by AI systems.

Legal constraints, operational risks and governance challenges continue to slow large-scale deployment of AI technologies.

Lessons from the Amazon Incidents

The outages at Amazon illustrate the complexities of integrating AI into critical operational systems.

While AI tools can accelerate software development and operational efficiency, they also require strong oversight, updated knowledge bases and reliable governance frameworks to avoid costly disruptions.

As companies invest heavily in automation and AI-driven productivity, maintaining a balance between machine intelligence and human supervision remains a crucial challenge for large technology organisations.