One of the foundational problems (and yes, there are efforts to improve this situation) with the much touted "Internet of Things" (#IoT) is that it can be complicated and unreliable.
Setups are often needlessly complex ("how many computer scientists does it take to configure a light bulb?").
Linkages between primary user interfaces (like #Google Home) and the third-party servers that actually control the devices are often fragile and under-provisioned -- and I blame those third-party firms, not Google) leading to Google saying it can't reach the service (when actually the command got through but wasn't acknowledged), or *actually* can't reach the service sporadically or completely for significant periods of time -- and often you find the third-party websites are also down and you can't get control that way either.
Just not ready for prime time.
Just one example: #HoneywellThermostats ...
@lauren it gets even worse when you factor in the distrust factor: e.g. you use VLANs to do things like prevent your cameras from accessing the internet and your Google devices from accessing pretty much anything that isn't the internet.
By flashing lightbulbs and smart switches to open source firmware, putting some MQTT glue on a server and writing a few short python scripts I got to the point where I could at least control lights and switches by saying vaguely natural sounding phrases to my iPhone.
Net cumulative effort savings: negative eleventy billion.
@lauren OTOH this was about the time Apple added the ability to do voice recognition entirely on the phone itself rather than delegating it to its servers, with the result none of the above uses the cloud.