There’s an S curve when it comes to embedded software implementation. The first steps of an implementation might be very easy and lead to a false sense of certainty and confidence. Then, you hit a snag. Pareto Principle sets in and you spend inordinate effort to implement the remainder of the project (this could be media intent handling for Alexa, for example). Then, things speed up again as you complete the implementation get back to things you know how to do.
Subsequent implementations take less time as you increase the slope of that middle line on the S curve. When you start to realize that the biggest challenge in setup was WiFi password issues, you can start to ask about this in the first troubleshooting steps.
The bottom right quadrant is where we get into trouble. Maybe it’s also the cause of the infamous Planning Fallacy cognitive bias. When we had first timed out the release of the Ubi, it was based on us being able to use Google Now on Android, even if there was no SDK. The unknown unknown to us was the inability to control the response from the device or trigger it outside of a hacky method (simulation a button push).
First time implementers of new voice services will likely face the same obstacles in learning of their unknowns.