Amazon plans 'deep dive' internal meeting to address AI-related outages

Amazon is preparing for an important internal meeting to tackle a series of recent outages, including those attributed to AI-assisted coding errors. Scheduled for Tuesday, this session, known as "This Week in Stores Tech" (TWiST), will be led by Dave Treadwell, a senior executive responsible for the technical infrastructure of Amazon's website. Treadwell informed employees that the upcoming meeting would focus on the root causes of these outages, stating, "This will be a deep dive into some of the issues that got us here." The meeting is set to commence at 12:30 p.m. ET. In a memo, he acknowledged the recent poor performance of the site and its related infrastructure, noting that the frequency of high-severity incidents (referred to as Sev 1s) has heightened the need for this discussion. In just one week, Amazon faced four high-severity incidents, prompting Treadwell to emphasize the importance of restoring the company's strong availability posture. A spokesperson for Amazon clarified that TWiST is a regular weekly review, where retail tech leaders assess the performance of store operations, including website and app availability as part of their commitment to ongoing improvement. The urgency of this meeting follows a significant malfunction last week, which left users unable to check out, access account details, or view product prices for approximately six hours. Amazon attributed these issues to a "software code deployment." As demand for AI services escalates, Amazon is investing heavily in infrastructure, anticipating $200 billion in capital expenditures this year—more than any of its competitors. However, this expansion comes alongside workforce reductions. Earlier this year, Amazon laid off about 16,000 corporate employees following previous cuts that eliminated approximately 14,000 roles. In a separate memo, Treadwell noted that AI-assisted changes have contributed to the incidents, citing "GenAI tools supplementing or accelerating production change instructions, leading to unsafe practices." He recognized that the necessary best practices for generative AI usage are still being developed. To address these challenges, Amazon plans to reinforce safeguards, including requiring higher-level engineers to review AI-assisted production changes initiated by junior staff. Treadwell expressed that temporary safety practices will be introduced to ensure greater control over essential retail operations while the company invests in more robust long-term solutions. While Amazon Web Services (AWS) has experienced several outages recently, the company clarified that these incidents are unrelated to the issues being discussed in the TWiST meeting. Notably, AWS faced a significant outage in December after changes were made by its Kiro AI coding tool, although Amazon stated that this incident was due to user error rather than AI involvement.