AI Transforms Contact Center Quality Assurance


The introduction of artificial intelligence (AI) technology into contact center quality assurance processes is nothing short of transformational. It is of particular value to healthcare contact centers, which must comply with high standards of performance and strict regulatory requirements.


But first, some background.


Since the inception of contact centers, there has been a traditional way of evaluating recorded calls and providing feedback—a member of the quality team randomly selects calls for a particular agent or program, listens to the recorded calls, scores them according to pre-established criteria, and discusses their observations with the representative. Often, the call is listened to together during the feedback session.


Most typically, the number of calls evaluated on a weekly or monthly basis is determined by the tenure and past performance of the representative.


This process, if executed effectively, has a valuable outcome: if the representative applies the feedback to future similar calls, those calls will then more closely adhere to the pre-established quality criteria.


There are both shortcomings and inefficiencies inherent in this process:


  • The pre-established criteria may not measure some of the more nuanced aspects of the calls
  • The feedback provided is specific to what transpired in the evaluated calls, and may not be relevant to all future calls
  • A very small percentage of the representative’s calls will be evaluated, even if their tenure and/performance put them at the high end of the requirements
  • It is highly dependent on the competence and objectivity of the evaluator
  • It is almost exclusively dependent on human labor, and is therefore expensive


AI technology changes all of that.


In the context of contact center quality assurance, AI is the ability to review ALL call recordings and process them through a platform that “reads” each recording and analyzes it according to a number of factors, including:


  • The “sentiment” of the customer (positive, negative, or neutral)
  • Talk-to-listen ratio (the percentage of the call that the representative was talking versus listening)
  • Average talking speed (words per minute spoken by the representative)
  • Average pause (the average length of pauses taken by the representative during the course of the call)
  • Identification of key words or phrases (those words or phrases that should drive a specific response or action on the part of the representative)


In addition to providing objective, quantitative reports on the identified factors, the value of the AI analysis is that it pinpoints calls—or sections of calls—that warrant scrutiny by a human being.


A few examples, tied to the factors listed above:


  • The exact moment the sentiment of the customer changed—e.g., from negative to neutral, from positive to negative, from neutral to positive. What happened in the call immediately prior to this change and what can be learned from it?
  • Calls in which the talk-to-listen ratio is outside of the typical range for the type of call being handled. Why was this and what impact did it have on customer sentiment?
  • Calls in which the average talking speed of the representative was higher than 150 words per minute, which may have made it difficult for the customer to absorb what was being said.
  • Every instance of a pause on the part of the representative that is longer than (for instance) two seconds, as longer pauses can negatively affect communication and customer sentiment.
  • Every instance of the use of a key word or phrase by the representative or customer. In a healthcare contact center setting, any word or phrase that signals a potential adverse event would be considered key (e.g., sick, ill, hospital). Was the appropriate action taken in a timely manner after the call?


Through key words and phrases, the evaluator can also check to see if any promised follow-up actions took place (e.g., I’ll call you; I’ll send you; I’ll email you; You will receive).


While there is great value in AI technology for the evaluation of individual calls, its most significant impact comes from how it enables the real-time gathering and processing of data across the entire contact center. It is an organic process—through the data created from thousands of interactions, the AI

continually learns more about what “success” looks like for various call types.


Insights improve over time as patterns develop.  Those insights are then integrated into training and refresher training.  The following are just a few of the insights gleaned from our recent experience with AI technology:


  • If a customer complains about an issue (for example, the cost of medication) their sentiment turns from negative to neutral or positive if the representative does much more listening than talking.
  • In inside sales team calls, the term “cheaper” evokes a negative sentiment in HCPs, while “affordable” is tied to a neutral or positive sentiment.
  • In patient engagement calls, a rate of speech between 140-150 words per minute yields a more positive sentiment than does a faster or slower rate of speech.
  • In concierge programs (where the representative guides the patient through every step of their treatment journey), patients respond positively to “supportive” phrases such as:
  • You have reached the right place
  • I am glad you called
  • I am here for you
  • I can help you with that
  • I will take all the needed next steps


All gathered insights are shared across the contact center, in both formal and informal training sessions, to elevate the quality of calls and the customer experience.


AI technology also plays a key role in the initial training of contact center representatives. The training content is informed by the insights gathered through the processing of calls through the AI platform. The sharing of actual calls and their associated analysis in training highlights these insights.


Another training approach is to script and create recordings that illustrate what “effective” and “ineffective” sound like for the call type (or types) the representatives will be handling. This is a creative way to show the extremes for each key factor, which likely don’t exist in any one “real” call.


It is important to emphasize that AI is a supplement to the efforts of the quality team, not a replacement for those efforts. It directs evaluators to listen to a targeted group of calls that have highly positive or highly negative customer sentiment, contain a key word or phrase, or have quantitative factors (e.g., talk-to-listen ratio) that are significantly outside the norm. The evaluation of these calls leads to learnings that will raise the quality level across the contact center.


As we all know, advancements in technology are developed and implemented continually and swiftly. As effective as the current AI technology is in ensuring quality and success in healthcare contact centers, the best is yet to come.


Denise Dixon

SVP Client Services

Diligent Health Solutions


Tim Abbott

Business Unit Director

Star Outico