Contact Center AI - Chapter 2

Companies have always had difficulty obtaining data from their customer service interactions. This data consists of items such as the content of the interaction and the agent performance. In the beginning, the information was not gathered. More recently, the data is being gathered and stored, but it is underutilized – firms do not know what to do with it. Traditionally this data has been siloed in the contact center platform and not available for analysis with data from other software such as CRMs. If this data were more accessible, then companies would be able to make better decisions and adapt to changes faster.  The landscape is changing as interaction data from both chatbots and agents can now be more easily gathered and analyzed to help improve the contact center. 

Once chatbots are deployed, a wealth of data can be gathered. The first level of data gathering is simply discovering why the customer initiated a chatbot session. Initially, this information can be used to supplement the knowledge base or FAQs. For example, a customer may have difficulty finding out how to optimize a workflow in their SaaS software from the website documentation. When the company receives many chatbot requests for this, they can prioritize this documentation update.  Additionally, investigating the chatbot content allows a firm to react more quickly to market changes. If there are many questions about the capabilities of a product line that the company does not have, they may wish to add this capability to the product roadmap to remain competitive. Looking deeper into a chatbot conversation, companies can now use NLP to identify customer sentiment and better understand how they are perceived. An added benefit is that sentiment can influence how a chatbot responds to a customer inquiry. 

As artificial intelligence becomes more accessible, vendors now offer speech analytics solutions with advanced capabilities. Speech analytics can provide data from voice calls that was previously unavailable. Two of the most popular outputs include call transcription and sentiment. When calls are transcribed, speech analytics can perform NLP on the output to identify key words and phrases. Businesses can now understand when important words like a competitor’s name or a product name arise in a conversation. Taking this a step further, business analysts can study these phrases over time to identify trends. When marketing analyzes these trending words and phrases, they may elect to create a new marketing campaign for an upstart competitor in their space who is being mentioned more frequently. If companies wish to take a more active approach, they can use an advanced query tool to dig deep into calls to identify the surrounding verbiage of these key words. This can provide them additional insight as to whether a competitor was mentioned positively or not. Sentiment data supplements the actual words themselves. By studying sentiment, supervisors can understand how an agent’s performance affects the mood of the caller. For example, if an agent offers a promotion to a caller, does this illicit a positive or negative sentiment? Sales management can then make appropriate changes to their promotion strategy based on these results. 

Now that voice and chatbot data is easier to obtain, it can be uploaded to a data warehouse. The real advantage comes when combining these insights with other data, such as data from the CRM and then feed it all into an analytics engine. Companies can not only get a better picture of a single interaction, but also obtain bigger insights that can substantially help increase sales, improve service levels and reduce customer effort.