Automated origination methods came under fire recently in Canada where the practice of using an automated method to assist in (or to perform) property valuation assessments for mortgages is being challenged.
The article claims that this automation method and a program (named Emily) has contributed to escalating home prices by using reference data instead of actual valuations. Some people might read that article and see automation itself as the danger.
When it comes to banking automation, I think there are multiple points to consider.
Done well, automation helps take some of the drudgery out of many banking processes. I enjoy the stories from former colleagues when I would be designing a new system and they would explain how things used to work when it was completely manual — interesting stories, but the manual work itself sounded tedious and error prone. Automated systems (designed well) help introduce checks and balances that help the bank make better decisions about which loans to fund, and to highlight risks that may not be apparent to the person at the counter. Automation helps to ensure consistency and accuracy.
Automating processes is also about increasing productivity – doing more with less resources.
Poorly designed automation processes can take a bad idea and amplify it — making it consistent and making sure that everyone follows it. Really bad ideas automated can create enough hubris that they go undetected just because of the complexity of the process and systems; the assumption is that if it’s automated, it must be a great idea. I’m not sure about Emily itself, but I have seen systems that were designed for one thing then used for another. The purpose of the initial design was a great idea, but the later use was a bad idea — so the automated system itself is not the only concern it’s also how that system is applied in practice that will determine whether it works or not.
I really wanted to follow the Clint Eastwood theme here, so I had to include something ugly. When I consider the ugly side of automation, I think there is not enough end to end thinking about the automated process. Automation itself is not a bad thing — but taking a bad idea and automating it doesn’t work. To me the ugly is that there is often not enough consideration of the total view of automation and how to weed out the bad ideas that are being amplified in the process.
Automation is not necessarily bad or good, much depends on how it is done and whether or not the full end to end process is considered or not. Avoiding bad automation requires design thinking and a perspective that considers to total picture.