In my previous blog, I introduced the concept of Intelligent Customer Management that we discussed at NexJ Client Day. Many of our Wealth Management and Private Banking customers were eager to discuss how it could explicitly apply to them. To apply the concept, we need to start with business goals. In Wealth Management, business goals often include growing assets under management and improving customer loyalty. How can Intelligent Customer Management help advisors accomplish these goals?

Previously, I wrote a blog on Customer Loyalty Programs. With a customer loyalty program, we match tiers of customers with interaction patterns that we think will result in increased business. Advisors are proactively prompted to perform the interactions at the right time. Without Artificial Intelligence (AI) and machine learning, the plans are very prescriptive and rely on assumptions of what works and doesn’t work based on the performance of top advisors. Typically they are effective but, with data and analytics, they can be optimized for even greater success.

First, we use unsupervised learning to divide our customer base into cohorts of like customers. This process is similar to tiering, but uses machine learning to group similar clients together based on correlations that would not be obvious in a manual process. The example we used on Client Day was high net worth frequent flyers that were close to retirement. It’s unlikely that a manual process would have identified a group of clients that will invest more when treated similarly, based on these attributes.

Now that we have the cohort, we need to determine the correct patterns of interactions that will lead to more investments. Another machine learning method, supervised learning, is used to train a decision model to make predictions or deliver a score. The decision model is trained by using historical data to find patterns. The data must include both inputs and outcomes. In this case, the inputs are all of the interactions that occurred for a given household, and the outcomes are the subsequent increase or decrease in assets over a prescribed time. The result is a decision model that predicts the best set of interactions for a given client.

The prediction can be further tuned if the inputs include information locked in unstructured meeting notes and emails. Natural Language Processing is used to extract names, relationships, intentions, and tone from the unstructured text. All of these attributes are added to the input to create better decision models.

The benefit for the advisor is a series of recommended interactions across their book of business.

Is your firm considering the use of AI to influence advisor behavior? I would love to hear your experiences.

Tune into my next blog where I will discuss Intelligent Customer Management in Commercial and Corporate Banking