Advanced AI topic clustering in Customer Service Insights

As the power ofartificial intelligence (AI)has evolved over the past few years, ithasshownits power to transform businessesto be more productive andtoserve customers better.

Using the power of AI, systems can digest large amountsof data,predict future patterns,and identify the best solutionsquicklywith minimum human effort. On the other hand, system capabilitiesthat rely onAI areusuallyhard to implementas theyrequire people who haveexpertise inboth AI and the business domain.

Thats why we bring youDynamics 365 Customer Service Insights. It focuses on your customer service scenarios andmakes itvery easyfor you to useAIalongside traditional business intelligence. This means you canmake better data-driven decisions, without requiringextensiveAI knowledge.

Using anout-of-the-boxnatural language understanding model, itgeneratesinsightson top ofyoursupportcases byautomatically groupingsimilar cases into topics.In addition tohigh-qualityAI models,it alsocomes withseveral advanced capabilitiesthat work togetherto provideyou withamarketleading solutionfor actionableinsights.

  • AI-discoveredtopicswith automaticoptimizationmakeitmore usablein real-world scenarios
    Thanks to its semantics-embedding model,Customer Service Insightsdoes a much better job togroup similar casesthan simplyplain text matching. Itunderstands thetext semantics whenevaluatingcasesimilarity,sothatcasesusingdifferent terminologiestodescribethe same thing can be grouped together.This capability helps to reducegeneratingduplicated topics.For example,casesthatdescribeissueswith the termspromotional certificate andcoupon code areconsidered asdescribingthesame 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 topicisassignedan auto-generatednamethat can represent itsrelevant cases.Once atopicname is assigned, either automaticallyby AIor by users who renametopicslater,Customer Service Insightstrackstopicsin the past to keeptheir namesstableand maintainable overtime.

  • Our AI model continuously improves fromcustomergestures
    Customer Service Insightscomes with a well-tuned AI model. However, we also understand one model cant fit alltypes of yourbusiness and needs.Thats why weaddedfeaturesthat allow you to manageandupdate your topicsbased on your needs. Furthermore, weenabledour AI model withacontinuousimprovement capabilitythat can learn andimproveitselffromgestures, such asusers givingthumbs-upor thumbs-down feedback,renaming topics,ormoving casesto another topic. With continuous use,youwill end up gettinganexclusiveAI model thatworks betterinyourspecific 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

  • Weprovidesimpleoptions foryou tocontrol AI results
    There are some other scenarioswhere you want to getbetter AI resultsbutitshard to be automated. Weve been trying to make those scenarios easier for you byproviding simpleconfigurableoptionsand settings. Changes tothose settings will guide the AI algorithmsto generate theresults that work better for your business needs,but again,without requiring youtoknowthe complexityofthose algorithms.For example,data cleaningisknown to be critical – butcomplicated – toimplement inthe AI world, as you can get very poor resultswhenyour datacontainsa lot ofirrelevant information.To solve that problem,Customer Service Insightsprovides data-cleaning settings, where you can get your data cleaned simply witha 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 ormore specific way.You can control the scope of how your topics are generated bysetting 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 getquestions from customerslike youwhoare interested inlearningmore details aboutthese AI capabilities,orwonderhowit works better than other solutions.

Thus,Im glad to introduceyoutoa seriesofAI inside-outblog poststhatwe are planningfor the next few months!

Youllseemonthly posts from our AIexpertsthatcover thekeyareassuch as what is clustering and why its better for topic discoverythan other technologies, what is semantics embedding andwhy itplays a critical rolein Customer Service Insights,howweenableduser-friendly AI controlsto addressreal-world clusteringscenarios, and howour AI modelscontinuouslyimprovewith usersgestures among others.

We will updatethisblog postwith the linkstootherpostsin this seriesonce they arepublished.Please stay tuned!Iftheres any other area you are interestedand would like to learn more, pleasefeel free to leave your commentsbelow.

As always, your feedback is critical for us to prioritize whats nextwithin the product. If you have any suggestions or ideas, please dont hesitate tosubmit an idea or vote on others ideas.

If you have questions aboutCustomer Service Insights, were always available at theCustomer Service Insights forumto help you.


To learn more, visit:

The post Advanced AI topic clustering in Customer Service Insights appeared first on Dynamics 365 Blog.

Comment List