We present IntraVox, a novel voice-based interaction system that supports users with complex needs by introducing a highly personalized, human voice command between smart home devices and sensors. Voice-enabled personal assistants (e.g., Amazon Alexa, Google Home) can be used to control smart home devices and increase the quality of life. However, people suffering from conditions such as dementia or learning disabilities and autism (LDA) can encounter verbal and memory problems in communicating with the assistants. IntraVox supports these users by verbally sending smart home control commands, based on the sensor data collected, to a voice-enabled personal assistant using a human voice. Over 12 months, we conducted 7 workshops and 9 interviews with 79 participants from multiple stakeholder categories-people with LDA, older and frail people, social and health care organisations. Results show that our solution has the potential to support people with complex needs and increase independence. Moreover, IntraVox could have the added value of reinforcing the learning of specific commands and making visible the inner workings of a home automation system, by giving users a clear cause-and-effect explanation of why certain events occur.