I just don't see a lot of utility going forward unless they liberalize the service and allow other people to develop new experiences for users. That should also give them new data to mine and allow good voice UI patterns to bubble up.
Not sure how they should go about designing a framework for developers to plug-in to, though. It's a tricky problem. But presumably they've learned some things after 'sharing' some of Siri with Uber and Facebook and that could help them move forward.
I do think that the walled-garden is concerning since you can't use it with all music services, for example, but on the other hand, if its a premium speaker system that might be enough to push me the other way and switch to Apple music. The question is is the product compelling enough to get people to buy into the platform.
In other words, a voice platform that works 10x better than another may only be slightly more useful to the average consumer while if the actual speaker is 10x better, it might be 10x more compelling.
I think part of the problem is that the things 'voice' is good for tend to have humans on the receiving end that then carry out complex, human actions. "Stu, could you schedule a dinner for me at a good local Italian place at 7:30 and invite Scott and Robin?"
Well shoot, that is a lot to unpack! It's a hell of a lot for a computer to unpack. It's a lot for anyone to unpack unless they know you and your circle of contacts. Carrying it out would require interaction with at least 3 people over a variety of mediums.
I suspect that by opening Siri up (And other digital 'assitants') it might promote the growth of infrastructure and services that would begin to make some of these more advanced queries a little more tractable.
To be really compelling, voice needs to offer the whole AI concierge experience. At a basic level, it needs to be able to deal with queries like "Find me a good place to stay in Barcelona" and "Find me a new jacket". It needs to ask questions as needed, to operate with the initiative to search tens or hundreds of sites, and to recognise the useful data in the results.
With Google, Amazon, and the rest, search is becoming more and more of a problem, not less. For non-trivial searches, finding good products and/or reliable information can be incredibly time-consuming.
So currently voice is a bandaid on top of search tools that aren't progressing much, and may even be regressing. Voice has to solve the search problem first before the recognition and context awareness problems really become important.
It looks as if the industry is to trying to do this the other way around. I'm not convinced that's going to work. It works up to a point, but the point isn't as advanced as users expect it to be, and the overall experience can be disappointing.