AWS unveiled new AI coding tools at its re:Invent conference, aiming to streamline software development. The centerpiece is the Kiro agent, designed to enhance automation in coding by handling tasks like bug triage and code management across multiple repositories. AWS claims Kiro can significantly reduce project timelines, citing a case where a project originally set for 18 months was completed in just 76 days with only six developers. However, developers must trust the technology, given the potential for errors that have plagued AI tools in the past. Alongside Kiro, AWS introduced a DevOps agent intended to relieve the workload of on-call engineers by providing continuous monitoring, incident triage, and performance improvement recommendations. This agent requires extensive access to applications and their components, which raises concerns about security and oversight. AWS reports a high success rate in identifying root causes of issues with its internal use of the agent. Users can restrict its access and interact through a dedicated Slack channel for better control. There’s skepticism around the effectiveness of these AI tools. A study indicated that AI agents fail to complete basic tasks 70% of the time. Many companies have adopted AI tools without seeing significant productivity gains, often requiring developers to spend time correcting AI errors instead. AWS is attempting to address these challenges, but the transition to greater AI autonomy in coding remains a significant risk.