Advanced AI topic clustering in Customer Service Insights

As the power of artificial intelligence (AI) has evolved over the past few years, it has shown its power to transform businesses to be more productive and to serve customers better. 

Using the power of AI, systems can digest large amounts of data, predict future patterns, and identify the best solutions quickly with minimum human effort. On the other hand, system capabilities that rely on AI are usually hard to implement as they require people who have expertise in both AI and the business domain. 

That’s why we bring you Dynamics 365 Customer Service Insights. It focuses on your customer service scenarios and makes it very easy for you to use AI alongside traditional business intelligence. This means you can make better data-driven decisions, without requiring extensive AI knowledge. 

Using an out-of-the-box natural language understanding model, igenerates insights on top of your support cases by automatically grouping similar cases into topicsIn addition to high-quality AI modelsit also comes with several advanced capabilities that work together to provide you with a marketleading solution for actionable insights 

  • AI-discovered topics with automatic optimization make it more usable in real-world scenarios
    Thanks to its semantics-embedding model, Customer Service Insights does a much better job to group similar cases than simply plain text matching. It understands the text semantics when evaluating case similarity, so that cases using different terminologies to describe the same thing can be grouped together. This capability helps to reduce generating duplicated topics. For example, cases that describe issues with the terms “promotional certificate” and “coupon code” are considered as describing the same type of issue.

    Customer Service Insights detects similar text semantics from case titles to reduce topic duplication

    Figure 1: Customer Service Insights detects similar text semantics from case titles to reduce topic duplication

    In addition, each topic is assigned an auto-generated name that can represent its relevant cases. Once a topic name is assigned, either automatically by AI or by users who rename topics later, Customer Service Insights tracks topics in the past to keep their names stable and maintainable over time.  

  • Our AI model continuously improves from customer gestures
    Customer Service Insights comes with a well-tuned AI model. However, we also understand one model can’t fit all types of your business and needs. That’s why we added features that allow you to manage and update your topics based on your needs. Furthermore, we enabled our AI model with a continuous improvement capability that can learn and improve itself from “gestures”, such as users giving thumbs-up or thumbs-down feedback, renaming topics, or moving cases to another topic. With continuous use, you will end up getting an exclusive AI model that works better in your specific business scenario.

    One way to improve AI mode is to provide thumb-up/down feedback to indicate if cases belong to a topic

    Figure 2: One way to improve AI mode is to provide thumbs-up or thumbs-down feedback to indicate if cases belong to a topic

  • We provide simple options for you to control AI results
    There are some other scenarios where you want to get better AI results but it’s hard to be automated. We’ve been trying to make those scenarios easier for you by providing simple configurable options and settings. Changes to those settings will guide the AI algorithms to generate the results that work better for your business needs, but again, without requiring you to know the complexity of those algorithms. For example, data cleaning is known to be critical – but complicated – to implement in the AI world, as you can get very poor results when your data contains a lot of irrelevant information. To solve that problem, Customer Service Insights provides data-cleaning settings, where you can get your data cleaned simply with a couple of clicks.

    Clean your case data by a couple of clicks

    Figure 3: clean your case data by a couple of clicks

    In addition, you may want to group topics for your customer service system in a more general or more specific way. You can control the scope of how your topics are generated by setting topic granularity in Customer Service Insights.

    Set the scope of topics based on your business needs

    Figure 4: set the scope of topics based on your business needs

While we make all capabilities easy to use in the product, we often get questions from customers like you whare interested in learning more details about these AI capabilitiesor wonder how it works better than other solutions.  

Thus, I’m glad to introduce you to a series of AI inside-out blog posts that we are planning for the next few months 

You’ll see monthly posts from our AI experts that cover the key areas such as what is clustering and why it’s better for topic discovery than other technologies, what is semantics embedding and why it plays a critical role in Customer Service Insightshow we enabled user-friendly AI controls to address real-world clustering scenarios, and how our AI modelcontinuously improve with users’ gestures among others. 

We will update this blog post with the links to other posts in this series once they are published. Please stay tuned! If there’s any other area you are interested and would like to learn more, please feel free to leave your comments below. 

As always, your feedback is critical for us to prioritize what’s next within the product. If you have any suggestions or ideas, please don’t hesitate to submit an idea or vote on others’ ideas. 

If you have questions about Customer Service Insights, we’re always available at the Customer Service Insights forum to help you. 




To learn more, visit: