Julia is the Chief Technology Officer at ABC Corp., which operates in a highly regulated industry. Last year under Julia’s leadership the technology department played a central role in driving several key operational efficiency improvement initiatives at the firm, saving the firm millions of dollars through automation. In their recent quarterly meet, Travis, the business unit head of one of their largest customer-facing businesses shares plans to further bring down costs significantly in his unit. He speaks with Julia at length about this and enthusiastically shares a very aggressive headcount reduction target in specific business processes. He tells Julia that he looks forward to seeing her continue to produce great results this year as well, leveraging their data science and automation teams. Julia senses Travis already setting high expectations especially on the data science team to leverage machine learning to increase the automation significantly. Julia agrees with Travis to come back with her overall strategy and plan in a couple of weeks. How does Julia approach this? Read on.
Julia is a staunch believer in “horses for courses” and “best of breed”. In this case she has practically seen many cases of what can and cannot be achieved through traditional automation versus automation/semi-automation through machine learning*.
She knows that last year she and her team achieved a great deal of process automation in several key processes already in this particular business unit. She explains to Travis that by nature machine learning deals with areas of uncertainty and continuous learning from data to keep making more accurate predictions. One cannot expect complete automation especially given that these specific business processes that Travis is talking about involve a great deal of reputational and regulatory risk with no tolerance to erroneous processing.
She further explains that in these specific parts of the system, her ML model will still drive significant cost reductions, serving as a digital assistant to the operations staff with predictive insights that will greatly reduce manual labour and human errors, thus making them more productive. She shares with Travis her overall strategy that comprises an upstream automation component followed by the ML Model, and an in-built continuous feedback loop to recommend refinements to the automation rules engine based on the learning over time. This is in addition to the “natural” continuous learning of the model from newer data, continuously increasing its accuracy. She provides a high level plan to deliver significant cost savings derived partly through rules-based automation and partly through machine learning, while managing the operational risks involved in these critical business processes.
* See also my blog on this topic!