When a user says "Okay Google" to a Google Home device, the device's microphone begins recording audio locally. The device then sends the audio clip to Google's servers, where it is analyzed by Google's speech recognition software to determine what the user said. Once the speech has been transcribed, the device then sends the text representation of the user's command to Google's servers, where it is interpreted and processed. Google Home devices are designed to only listen for the "Okay Google" wake phrase, and the microphone is only activated when this phrase is detected. This is done to protect user privacy and prevent the device from recording audio when it is not intended to. Once the user's command has been processed by Google's servers, the device will respond to the user's request with the appropriate action. For example, if the user requested the weather, the device will provide a weather forecast. If the user requested a song, the device will play the requested song. In summary, "Okay Google" works on a Google Home device by activating the device's microphone locally, sending the audio clip to Google's servers for transcription and interpretation, and then responding to the user's request with the appropriate action.