Agentic AI is here. Is your bank’s frontline team ready?
McKinsey Insights
Summary: This article discusses how agentic AI can be utalized in financial services operaitons. The authors argue that banking operations are particularly well suited for AI adoption due to the high volume of repetitive tasks and data entry. They identify five key essential uses of agentic AI.
Thoughts: Reading this article is indicative of how AI is currently being utilized in the economy. The potential of AI is enormous, but only technically advanced businesses are able to capture this potential; even fewer are capturing the potential of generative AI. Writings on this topic is therefore rarely insightful and often even less actionable. McKinsey Insights suffer especially from this, compared to how they fare in other domains. Although, this is probably the most insightful article by McKinsey on AI that I have read, perhaps even the most insightful from any business magazine. I enjoyed how they argued that banking operations are particularly well suited and motivated for the application of AI. It seems like this approach can be applied to other industries as well, which is interesting. It is also interesting that the article discusses only speed of operations, not quality of service. When discussing data entry, the authors can get away with this, but that is not always the case. We see an example of this in the final section on pricing, where the authors argue that AI removes the need to rely on guesswork by replacing it with analysis. I read this as fundamentally wrong. AI is not interchangeable with analytics; that is a common misconception. The use of an AI model has to be backed by a hypothesis of causality. The authors state that bankers currently feel they are relying on guesswork, but if they do not have a method now, AI will not conjure up new causal relationships. Bankers may feel more confident because they believe their results are supported by analytics, but in reality the guesswork has merely been shifted to an arbitrary AI model. This is not a desirable outcome. In other words, institutions need to develop a theory of causality before they can meaningfully 'use AI'. But the other points they make are solid. Most of the essential uses of AI, like Lead Nurturing and Prospecting are actually great use-cases.