Artificial intelligence (AI) is a trending topic in the communications sector. Businesses often assume they don’t have the financial resources to utilize AI and machine learning in-house. In our webinar with Xprime.ia and Encore Capital Group, we reviewed the importance of artificial intelligence and machine learning and how companies have more capability than they think to create predictive models. Using artificial intelligence to evaluate and understand past contact center conversations shows businesses what to expect and gives them the opportunity to be proactive in the future.
Here are the key takeaways from our presenters. Attend the webinar replay on February 27th featuring a Q&A web chat with Rick Britt.
Sheraz Shere, Founder of Xprime.ia, Artificial Intelligence & Machine Learning Consultant at Google
Xprime.ia is an independent consulting firm focused on building machine learning and artificial intelligence proof of concepts. Xprime was founded to democratize artificial intelligence because many companies felt doing pilot machining and AI was privy to big fortune 100 companies that can afford to pay large technology companies to do the research and work. Sheraz believes that all businesses can do lightweight experimentation and build machine learning models. The most challenging part in achieving AI success is finding the talent or the people that can do these activities. Xprime helps businesses that don’t have the in-house staff manage their AI strategies.
3 Driving Factors Behind AI
- Big data is a term we hear often, and data is the fuel for AI. We live in a digital world where data is captured everywhere, and businesses have valuable information about customers and their expectations. Artificial intelligence gives businesses a way to decipher this information.
- Major computational improvements. Buying computational power is no longer expensive, making it accessible to more businesses regardless of size.
- Open Source revolution. The tools used to machine learning and AI are not proprietary meaning they are available to any business that wants to build and train machine learning models.
If contact centers had a human listening to every call, what would they be able to learn and how could they use that to predict what a customer is going to do next? Customer audio and contact center interaction are two of the largest dark data centers out there meaning they are untapped.
By using machine learning with speech analytics data and transcripts, businesses can now interpret past calls, build a predictive model, and uncover what steps a customer will take next.
For instance, businesses can predict if customers are open to cross-sells, attrition, understand call topics, and determine patterns causing calls.
Rick Britt, Vice President of Artificial Intelligence, CallMiner
CallMiner has been in the AI business for a while, a good long while, it has just not been transparent to our clients. The basis of speech recognition is a neural network acoustic and language model. Tools used by our customers’ analyst every day to analyze their customer interaction leverage AI techniques. These include Search QA, which leverages machine learning based statistical phrase pattern engine and TopicMiner® for auto topic discovery which leverages NLP vector clustering to identify trends. On this foundation of AI, we are poised to do what humans struggle to do. Complex pattern recognition. In this amazing 1 trillion word set, are all the patterns in each interaction, across every interaction, are the optimal interaction paths for industries such as healthcare, to ensure accurate resolution. Why do consumers call back or cancel a product? The answer is based on every interaction they have had. Or more complex question: who are our customers? What if you could predict who is the perfect person to interact with a customer, the list goes on. The patterns to identify these questions and answers are so close to impossible to see by humans, that it takes artificial intelligence to recognize them.
This is what we are doing, building the complex mathematics and models to find those patterns, no matter how small, to predict the things that are relevant to every interaction, including yours.
So what’s next for AI? 8 ideas
- Auto-Correlation: Identify language, metadata, behaviors that drive an outcome
- Auto-Predict: Surface a likely outcome based on characteristics of a contact
- Auto-Disposition: Based on historic evidence, automatically classify new contacts
- Auto-Score: Generic and publicly available AI predictive models for call scoring
- Auto-Redaction: Entity based extraction of vulnerable data
- Auto-QA: Measurement & prescriptive analytics to drive agent behavior
- Emotional Acoustic Scoring: ML models for classification of acoustic based emotions
- Semantic Discovery: ID of call topics and Summarization thru NLP and AI techniques
Shawn Feaser, Senior Manager of Data Analytics, Encore Capital Group
Encore Capital Group is one of the largest debt buying companies in the world as well as a CallMiner Eureka customer that started their interaction analytics journey 3 ½ years ago. They manage 80,000-100,000 calls per day.
When they started with interaction analytics, they looked at the favorite items like understanding silence and unproductive agent behaviors. They changed their viewpoint and decided to rebuild their analytics machine after looking at the events of the call as the DNA of interactions. They looked at each event in the call to understand the cause and effect of what happens next.
By providing structure and categorizing the data, Encore could make it usable. They provide data to the operations team in an easy-to-understand model so that they could make positive changes behaviors and processes on approaches to future interactions.
Now they can uncover more information on the ability of the call center agents by reviewing the calls and capturing the events where a supervisor is brought on the phone. Do certain agents rely heavily on the supervisor? What caused the need for a supervisor? In addition to reviewing the effectiveness of the agent, Encore can now use the data to measure the effort and ability of supervisors in managing customers and how well they can train their agents.
They’ve had so much success to date, that they are expanding their use of artificial intelligence and are in the process of building training bot call simulator. They capture snippets of multiple calls and provide different ways agents respond to those situations. New agents review the training examples and then use the bot to practice. Through the bot, they record responses to difficult call conversations and the bot provides an evaluation on how well they managed the situation. The goal of the bot is to shorten the learning curve and give agents with the right tools to handle a variety of complex situations.
By leveraging the info they have, Encore can share the information with other departments and gather ideas on how they can use the information to build a better customer experience. The goal is eventually add in other forms of correspondence, so every aspect of the customer journey is considered when creating new procedures and policies.