As AI technology becomes pervasive and starts to impact all aspects of the work done by businesses, managers can no longer avoid to face the choice on how and in what scope and capacity should they be using AI. While some tasks are more suitable for automation some might be better suited to humans working alongside and collaborating with the AI outcomes.
Before I continue with the post it is good to understand the difference between automation vs. augmentation. Automation is when the machine takes complete control of the task being done, for example if you wanted to automate credit or loan approval, machine can ask the users to input the necessary information and then execute models behind the scene to assess the expected profits vs. expected losses of approval and then decide to give an answer in real time. In the same scenario augmentation would entail users entering the input, machine does the computation and then credit officer makes the final decision based on some subjective criteria.
Books on the subject have often cited augmentation as the way to go and this could very well be linked to the fear of machines replacing human workers resulting in loss of jobs. The answer though is more complex than that as automation and augmentation are interdependent over time and changing scope of work.
There are couple of reasons why certain tasks might be better suited to augmentation. If there is subjective conceptualisation required in defining the problem statement, if there are not enough data examples for the machine learning algorithm to handle all types of exceptions and scenarios, if the current sophistication of algorithms is not able to handle the given problem with the desired accuracy, if the task is highly critical and sensitive (think of domains related to HR and healthcare) then complete automation might not make sense. However, this might change over time, for example a task that could not be automated today due to lack of enough data, could become a good candidate for automation as the machine learns new insights and patterns over time from new data that comes in.
Also, a task that has been automated today might change and evolve and hence be better suited to augment human intelligence instead of being fully automated. Consider how the role of a chatbot might change as a new and more complex insurance product is introduced and hence the customer queries can no longer be handled as well as they were in the past when there were fewer products being offered by the company.
The diffusion of AI across industries will benefit from management research on the topic as managers can no longer decide to sit on the sidelines of the debate. I am intrigued by the roadblocks to adoption by small and big businesses alike and one of the reasons is lack of managerial understanding about what can AI do for them and how should they use it. Automation vs. augmentation is one such area where we need to drive more clarity to enhance adoption of AI!
References
- Artificial Intleigence and Management: The automation- augmentation paradox by Sebastion Raisch, Sebastion Krakowshi
Comments