Just about anything Google or Amazon is offering is at best using their local devices as a technology bridge and at worst, talking to both your Google/Amazon device and to individual IoT hubs and stuff usually through the cloud associated with that device. In both cases, the logic and cross-IoT interaction is happening in the cloud often with multiple cloud hosts involved.
Until you’re using Home Assistant, Homeseer, openHAB, or some software that can talk directly to an internal IP address and API of the device, through a power line interface, or to a local hub for proprietary protocols (e.g. to shades or whatever), it’s probably going to have the cloud as an intermediary.
Then, you’re back to being impacted by internet outages, mergers and acquisitions, introduction or increase in subscription costs, and the whims of company’s marketing departments to deprecate systems or re-insinuate control if someone managed to reverse engineer it for local access.
In addition, each of these devices can directly become a risk for vulnerabilities and potentially lateral infiltration of your network. Unless you’re running a firewall and care to review logs daily, it’s hard to know what they’re doing behind your back.
If I can’t talk to the device or at least the hub that controls them directly in-house, I try to avoid it to ensure independence, cost control, and to reduce the perimeter that has to be managed for security.
There are lots of good points here about preventing the annoyance to start with: better fixture/bulb/aiming, timer. Are you willing to cede control at a time when you may be expecting someone or put it on briefly so you can see?
I recommendcthese other approaches first. Why make your neighbor do something to remedy the problem?