Amazon has widely implemented an internal tool called MeshClaw, which is capable of creating AI agents to carry out various tasks. Under the immense pressure of KPIs, some employees, in an attempt to showcase high-frequency AI usage to the management, leverage MeshClaw to automatically generate a vast number of superfluous AI activities. Their aim is to ramp up token consumption. Despite the company's assurance that the relevant data is not employed for performance evaluations, employees remain skeptical, believing that management is covertly keeping tabs on it. Moreover, employees express concerns regarding the tool's requirement for access to user operation permissions to execute tasks, as this could potentially give rise to security risks, such as errors caused by AI agents.
