😬 AI agents turned out to be not the most pleasant colleagues - Google added its agent to the chats of real teams for five months. The agent participated in more than 20 work chats, got access to documents and a bug tracker, could write to employees, tag them and give reactions. But with social signals everything turned out to be more complicated. The AI ​​sometimes mistook heated discussions for conflicts, did not understand sarcasm, and even reprimanded colleagues for making jokes. As a result, some employees began creating separate chats without an agent, especially for sensitive discussions. People were uncomfortable with the feeling that the AI ​​was constantly present and “listening.” Source