> I think be the end of the desktop paradigm.
Think about how today's businesses generally use their employees: let's not beat about the bush: they're using Microsoft Office for producing documents and spreadsheets, doing research with Chrome, and (if applicable) creating business value using either line-of-business software ("Bob's Car Repair shop database in MS Access") or specialised content-creation software (Photoshop, AutoCAD).
None of those use-cases can be fulfilled by recent interaction paradigms (i.e. smartphones, tablets, or voice-control). VR's useful for visualisation, but that's a very limited use-case, and VR has been used for architectural and civil-engineering visualization for decades anyway (e.g. https://www.youtube.com/watch?v=T2CYLlSn1gA ).
Granted, when considering an iPad paired with a keyboard - we see some of those use-cases become almost workable again (namely word-processing), but it's essentially the same thing as a traditional laptop - just with considerably more painful precision object manipulation due to the lack of a mouse cursor.
I feel the biggest paradigm-shift over the past 10 years is the movement of database business applications away from Access, VB6, Delphi, etc to be web-applications, usually hosted locally, and more recently to being moved to multi-tenancy, subscription-based services ("the cloud") - this has now become acceptable to businesses: the trade-off of no-longer being in exclusive control of their data is increased uptime, reliability, and significantly reduced start-up and (usually) running costs - and being browser-based this has meant a shift away from WIMP - where an activity is contained within a small-ish, often fixed-size, "window", to a "page", which has its own interaction model (e.g. no more modal windows because you can always open a new browser tab).
The next iteration of human/computer interaction we'll see that impacts our (society's) actual working lives will be something that enables me to be more expressive without resigning any degree of control over what I produce - and right now nothing does that better than a large display which lets me interact with multiple applications simultaneously, a keyboard with which I can enter commands or prose almost as quickly as I can think it (with a very low error ratio too), and a mouse for quick, precise, (gorilla-arm-syndrome-free) content/command selection.
Until something comes along that enables me to be more productive for my employer the desktop interaction model will still reign.