Imagine, if you will, an organization that has to do some computing. Some data is gathered, it is processed by """machine learning models""", and some actions are taken based on the results of said processing. Those pesky machine learning models are a pain to develop, train and most importantly run. Who has the time and energy to deal with that? Enter some glossy tarpit with a GUI like Azure ML Studio, where everything is possible but nothing is particularly easy. At least the MBAs are happy, because there are no nerds cawing about CUDA versions and load balancing.
But wait!
For reasons of saying yes too quickly to people who should actually be told that their 'business logic' will be replaced by creative new forms of haruspicy — fortune telling by means of inspecting animal entrails — some pre- and post-processing rules have to be applied to the data. No biggie, we'll write a wrapper... until said priests of the business temple want to control the logic themselves.
I wish I was joking but I recently learned of the existence, somewhere in an undisclosed location, of an in-house data transformation tool built using draw.io. If you're unfamiliar: it's a tool for drawing diagrams. Users draw boxes and arrows, put some text in the boxes, and save their diagram pictures to some fileshare. The tool interprets these, transforms them into a data transformation DAG, which is then deployed. If you are of the school of thought that Excel-driven business processes are to be avoided, you will now probably experience some level of discomfort.
Every day we stray further from God's light.