In one embodiment, a computing system may receive input data from several artificial-reality systems used by participants to participate in a virtual meeting. The input data may include audio data and sensor data corresponding to the participants while they are in the virtual meeting. The computing system may determine meeting characteristics based on input data. The meeting characteristics may include individual behavioral characteristics for each participant and communal characteristics associated with the participants as a group. The computing system may determine an action to improve a meeting characteristic according to a predetermined criteria and generate an instruction for performing the action. The computing system may send the instruction to one or more of the artificial-reality systems used by the participants. The instructions may be configured to cause an application running on each of the artificial-reality systems that receives the instruction to perform the action.