Amazon pushes AI use and closely tracks adoption, even as some employees push back

Amazon is meticulously monitoring the integration of artificial intelligence within its retail division, referred to as "Stores," while encountering some resistance from its workforce. An internal document obtained by Business Insider reveals that the company is analyzing how frequently software engineers utilize AI technologies and the resulting impact on productivity. The initiative aims to enhance the software release pace across over 2,100 engineering teams by implementing what Amazon describes as "AI-native" methodologies. A select group of at least 25 teams is tasked with achieving a tenfold increase in output by the end of the year. The progress of these objectives is closely overseen by Amazon's senior leadership team, known as the S-Team. Generative AI is making waves in software development, significantly boosting coding capabilities recently, driven by tools like Anthropic's Claude Code and OpenAI's Codex. This surge in software production has also kept code quality relatively stable, prompting tech firms to capitalize on potential productivity enhancements. Amazon is pushing for an expansive adoption of its AI tools, a move that has sparked internal friction. CEO Andy Jassy has previously emphasized the importance of embracing AI technologies, warning employees to adapt or risk job loss. An internal document stated, "Treat AI like any automation investment. Actively seek opportunities to apply it, measure successes, and establish habits around those victories." This document, classified as "Amazon confidential," was created by a team dedicated to evaluating and enhancing AI tools utilized by thousands of engineers within the extensive retail organization. The company promotes open feedback to identify and address challenges early in the process, demonstrating a willingness to explore the potential of AI in the workplace. The document outlines a strategy to measure deployment rates and AI engagement while remaining cautious of "Goodhart's Law," which suggests that when a measure becomes a target, it may lose its effectiveness. In essence, as metrics evolve, human behavior adapts, potentially skewing results. Montana MacLachlan, an Amazon spokesperson, remarked that this initiative exemplifies the company’s commitment to investing in employee training and AI tool adoption. According to MacLachlan, the most significant improvements come when AI is integrated throughout the development lifecycle rather than as an afterthought. Currently, Amazon anticipates that 80% of its retail engineering teams will adopt AI-native practices, with approximately 60% already on board as of February. The utilization of Amazon's proprietary AI tools is also on the rise, with the AI Teammate—a Slack-integrated assistant—now active across more than 700 teams. Despite these advancements, the shift towards a more AI-centric culture has led to internal friction. Employees have expressed concerns regarding the top-down approach, overlaps in AI initiatives among teams, and the complexity of onboarding certain tools. Some staff seek clearer guidelines, while others desire more freedom to experiment with AI solutions. In response to feedback, Amazon is adapting its strategy. As of February, leadership intends to promote "collaborative AI practices" instead of mandating specific tools. The company is also enhancing its approach by automating metrics collection and allowing teams greater flexibility in AI adoption. The document emphasizes the need to "remove friction" and celebrate early successes to foster momentum. MacLachlan reiterated that Amazon does not impose a central mandate for AI tool usage, instead allowing teams to select the solutions that best suit their needs. The company's internal environment encourages lively discussions, highlighting its culture of innovation rather than resistance to adopting AI. Ultimately, Amazon's strategy balances ambitious AI goals with a pragmatic approach, focusing on practical solutions and value creation. Engineers are encouraged to experiment with AI tools to streamline their workflows, while managers are tasked with providing clear guidelines and easy access to these technologies. The document states, "Make AI tools part of your daily workflow, not something you reach for occasionally."