Amazon employees are automating unnecessary tasks using its internal AI tool to show managers they are using the technology more frequently, the Financial Times has reported.
The retail giant has started to roll out its internal MeshClaw AI in recent weeks, allowing users to create agents that connect to workplace software and carry out tasks on their behalf, the paper reported, citing three people familiar with the matter.
Some employees said that their colleagues were using the software to automate unnecessary activity to increase their token consumption in response to company pressure. Amazon introduced targets for over 80 per cent of its developers to use AI each week and began tracking token consumption on internal leaderboards earlier in the year.
Amazon told employees that these leaderboards would not be used in performance evaluations, but several staff told the paper that they believed managers were reviewing token usage statistics.
“Managers are looking at it,” an employee told the FT. “When they track usage it creates perverse incentives and some people are very competitive about it.”
The company recently limited access to AI usage statistics so they could only be viewed by employees themselves and their managers; previously team-wide statistics were posted internally. A person familiar with the matter told the paper that managers are discouraged from using token use to measure performance.
MeshClaw itself is inspired by OpenClaw, a viral AI agent creator released in November. MeshClaw can initiate code deployments, triage emails and interact with apps including Slack, according to people familiar with the matter.
In a statement to the paper, Amazon said that the tool enabled “thousands of Amazonians to automate repetitive tasks each day,” adding that it is “committed to the safe, secure and responsible development and deployment of generative AI for our customers”. However, multiple employees told the FT they were concerned over the security risks posed by agents that are allowed to act on a user’s behalf.
Earlier in the year, Amazon organised an internal meeting to investigate a “trend of incidents” related to AI-based coding after a series of website outages.
