Better consideration of voice requests