Business rules as preparation for Machine Learning
In Chapter 1, we suggested combining the two AI approaches (rules and ML) to build a self-driving car. The ML approach is great for fuzzier requirements (to identify whether it is a dog or a child standing in the road). Rules are better for requirements we can state clearly and that must always be implemented (for example, swerve the car to avoid a child but do not swerve for an animal as it risks a more serious accident).
Your organization should be able to follow a similar approach—there are rules that can be clearly written by a human expert (for example, buyers must have a 20% deposit for their home loan). And there are experiences that are harder to express—a senior bank official might have a feeling that a loan application is fraudulent and need further investigation, but might struggle to explain exactly why. In our business, we are likely to need both rules and Machine learning approaches to mimic both these...
