The present disclosure relates to processing tasks offloaded from one or more user devices to a cloud service. An example method generally includes receiving a command to be processed by the cloud service from a first device of a group of user devices. The cloud service generates a response based on the command. The cloud service determines weighting scores for each device in the group of user devices for each of audio output, video output, and haptic feedback based on context information about the group of user devices. The cloud service selects a plurality of user devices to receive at least part of the response based on the determined weighting scores, partitions the response into portions corresponding to the selected plurality of user devices, and instructs each of the selected plurality of user devices to output the one or more portions corresponding to the selected plurality of user devices.