What it was supposed to do and why it didn't do that very well
For me the key feature was the ability to plug in robots to integrate with other systems and services. This is what made it different from mail/IM etc. It could have enabled all sorts of commercial scenarios with different ERP systems having their own robots participating along with ERP users, suppliers and customers.
In that sense it was more of a collaborative, interactive workflow system than a fancy email system.
The reason it failed in this context is that the developer experience was awful. Poor documentation, non-existent test servers, lack of "local wave servers" for devs, unpredictable changes with little or no warning. The list goes on.
Compare this to the developer experience around force.com or Windows Azure. Both of these companies understand what it takes to develop good quality systems in the real world. If you buy into Google, you are buying a support nightmare. Who wants that?