Sure, but... are any of us actually confused by the statement that he's building something comparable to "Jarvis" from Iron Man? When I read that, I assume that means he's basically trying to make a Siri-like "AI" that can be a little more helpful.
I've put a fair amount of thought into this in the past, and it seems to me there are various things a Siri-alike could do, if hooked up and programmed to do it. For example, based on travel patterns, it could note that you almost always go home after work at around 6:30pm. Through the GPS in your device, it could tell when you're close to home, and automatically set a series of conditions for your arrival-- including things like lighting level, music, temperature control, etc. There are a bunch of things like this that require coordinating across devices, which I don't think Siri/Echo really do yet.
Another thing that doesn't seem like it would be too hard to do would be to make it more proactive. Theoretically let's say it has an appointment in your calendar that's 2 hours away. It has access to traffic information for the travel from one location to another, and sees that there's a half hour delay. So about 2.5 hours before the appointment, it automatically gives you a heads-up. "You'd better leave now, or you're going to be late." Or maybe it could monitor your washing machine and say, "You left some clothes in the washer a couple of days ago. You should probably move them to the dryer to avoid getting moldy and gross." These would have to be designed very carefully to make them useful instead of being annoying.
It could also possibly do things for you. Like say, "Hey, your mother's birthday is coming up in 2 weeks, and you usually order her flowers. Would you like me to order her the same bouquet as last year?" There are various possible problems with this, from having bug cause the AI to spend more money than intended, to the social aspects of having some of these things done automatically.
So what I'm getting at here is that there's a lot of stuff that these "AI" systems could potentially do, but we don't have them do because there are loads of potential problems, and it needs to be tested first. In some ways, I think it makes sense to start by having a rich tech guy pay for his own prototype, and figure out what he finds useful vs. annoying. It seems like a reasonable way to work out some of the kinks before you even get to normal testing.