It seems like every discussion today around the future of Support Services, Customer Experience, or Self-Service includes a component of Machine Learning (ML) and how it will revolutionize the engagement model. The potential use of these emerging technologies is endless and with the ever-increasing availability of ML as a Service, the capabilities are more and more within reach of a wide audience. However, in many of our discussions with Consortium members, services organizations are not sure how to apply these technologies or where to start.
We have found a simple, visual model is a good way to begin thinking about what it takes to achieve a tangible result, and have organized this work under the Predictive Customer Engagement initiative. It’s based on a double loop system made up of three main components:
- Analysis and Rules
- The Event Loop
- The Improve Loop
This post will focus on the Analysis and Rules ‘engine’ which is a critical part of the Event Loop.
Analysis and Rules
The Analysis & Rules engine is where we speak the most about applying the emerging Machine Learning capabilities. A four-tier model has emerged that we often use to help organize our approach. The model consists of:
1. Data Repository: Do you know what data you actually have? How is it stored, who controls access, and how clean is it? Without data, applying any machine learning to achieve an outcome is a wasted effort. In our discussion with many data scientists, up to 80% of their time may be spent on the data questions.
2. Machine Learning Technology: There are, what seems like, countless tools, techniques, and skills that can be used to create a trained machine learning model. Depending on the desired outcome different ones will be used, and this is where your data scientist is critical in helping use the right techniques for the right outcomes.
3. Capabilities: What are the analysis abilities that you are trying to achieve? Are you looking to uncover patterns, make predictions, optimize a system? Thinking about the capability needed to achieve the desired outcome will help narrow the focus of the effort.
4. Visualizations: Like anything, there needs to be thought into the best way the output will be consumed by the intended audience. Do you need dashboard, a chatbot interface, augmented reality, or even consumable outputs for a machine?
It is important to note that this is not linear thinking. All of the components need to be thought through and discussed to build an effective model. Iterations will be encountered as each layer is explored, and the impacts of one decision drive changes in all the other layers.
Along with this four-tier model, the work can be ‘Directed’ or ‘Emergent’. In directed work, we start with a goal and then move through the model to define the data we need to collect in order to achieve our desired outcome. In an emergent model, we start with the data we have and ask the question, “What can we learn from the data we have already?” Most companies are sitting on vast amounts of data that can be used and letting an innovative team play with the data can have tremendous impacts.
While this tiered model certainly does not capture the complete nature of how Machine Learning can be applied, or the complexity that can occur in building out the system, it is an effective way to start a discussion on how to approach these emerging technologies.