Approaching cognitive automation testing

As we continue through the automation life cycle once we have developed our automation, we can continue to test what was developed. While adding a cognitive component (Document Understanding, machine learning (ML) skills, and so on) brings additional complexities to automation from traditional rules-based robotic process automation (RPA), testing cognitive automation isn't too different from the traditional approach to testing automation.

In Chapter 6, Understanding Your Tools, we saw that when developing cognitive automation, we were essentially developing two distinct components: RPA and cognitive automation—we would build the ML skill first, then build RPA to leverage that ML skill. With this mindset, we can think of RPA and cognitive automation as modular components of the broader automation project, mimicking the design principle called separation of concerns (SoC).

SoC

This is a software design pattern where we separate...