How Machine Learning and A.I. are Helping Businesses Maximize their Value from Customer Support Centers

Establishing high-quality customer support is critical to fostering good customer experience. This is especially important for companies sparring for market shares in competitive industries like logistics, hospitality, telecoms, and banking. Consistently great customer experience can give businesses a leg up, while a single case of negative customer experience can damage the brand. In fact, customer experience has been speculated to overtake product and price as the key brand differentiator by 2020. Recognizing this trend, businesses look for ways to improve the quality of customer service without incurring additional labor costs. In this article, I want to discuss how Machine Learning and A.I. are applied in customer service. My goal is to illustrate how humans will interact with computers in skilled labor and highlight the benefits of these technologies as we continue down the path of automation revolution.

“We automated away 4 million manufacturing jobs in Michigan, Ohio, Pennsylvania and Wisconsin. And we’re about to do the same thing to millions of retail jobs, call center jobs, fast food jobs, truck driving jobs.” — Politician

First, let’s look at a bit of history on customer support center technology:

Brief History of Customer Support Technology

Call centers have come a long way since the days of human operators picking up phones and flipping through manuals. Today, they are more appropriately recognized as customer support centers that cover multiple channels like web, social media, email, livechat, and the traditional phone, at all hours of the day. Massive amounts of information coming into the systems are processed using state-of-the-art technologies to ensure as much useful data (and metadata) is captured as possible. In the next phase of innovation, industry leaders are using algorithms trained on this pool of data to create A.I. assistant robots that mimic and assist human agents perform their tasks. This is reflected in a couple of ways:

  • Automate informational and common transactional tasks —bots can perform repetitive tasks and create a self-serve experience for the customers. Majority of the customers today are resourceful and try to self-serve, meaning they will use the tools available to them (like websites, mobile apps, and chatbots) to resolve issues before asking for support. After all, the ability to self-serve on-demand gives customers considerable speed and convenience. This is desirable by the customers, provided that bots can respond to customer requests appropriately in a natural way. This is made possible by the advancements in computer’s ability to understand and generate human language (a.k.a Natural Language Processing). Once the bot correctly identifies the reason of the request, it’s then mapped to the corresponding response. This mapping is the product of a machine learning process that uses historical customer-agent records as the training data (a.k.a Supervised Learning). “When a certain question was asked, what did the human agents say?” The bot could then learn to imitate those customer support agents. Some customers might not even realize they are interacting with a bot. We’ve also seen evidence that show some customers seem to feel more at ease typing in questions and answers as oppose to conversing with a stranger. The trend shows robotic process automation will bring more (and increasingly complex) items on the self-serve menu and human agents play a critical role in the “training” and developing process.
Automated Response Flow
  • Help human agents solve complex inquiries — customer support agents get increasingly difficult issues and frankly can struggle if they are not well-versed in the topic. Beyond rapid information retrieval, machine learning algorithms calculate, filter, rank, and provide agents the most relevant information and suggestions that progress towards resolution. Whether or not the matching finds the exact solution, the algorithms make recommendations for the next best step (e.g. link to useful reading material). This is particularly useful for junior agents that need extra level of guidance. Taking a step further, this guided (or scripted) conversation can incorporate company protocols that take into consideration customer sentiment (via Sentiment Analysis) in dealing with potentially “abusive” calls. It’s up to the agents’ expertise to convey the information in a format that is easy to understand for the clients. Sophisticated bots use a Reinforced Learning approach to train by rewarding positive customer feedbacks (e.g. positive sentiments) and penalizing negative ones. Ultimately, the bot can learn the desired response that maximizes total reward (i.e. ones that deliver fast, easy service and drive customer loyalty). Having a robot assistant helps human agents tackle customer requests effectively and efficiently.
A.I. Assistant Response Flow
  • Generate predictive insights and analytics — machine learning models can be used to make predictions about the customer that benefit the agent. Using what the business already knows about existing customers, algorithms can provide insights to customer behavior like the contexts and timing of requests. This type of forecast can take different forms:
  1. Targeting and personalization methods that “profile” a customer with alike customers that exhibit similar properties, e.g. Look-Alike Modeling (which can be a classification or clustering model). Customers in the samecategory are expected to have similar experiences. Agents can get ahead of future issues and prevent them from happening (e.g. providing resolution to other affected customers in the category via mass email).
  2. Further, prognosis can be made about the CSAT (customer satisfaction) to give agents a better understanding of their interaction with the customer and the ability to make adjustments if needed. Collaborative Filtering can make predictions about the customer interests by collecting preferences from other customers. Having these insights gives the agents more control when interacting with customers because they are more certain about the outcomes. If needed, they can employ strategies that work (with other customers) to turn a negative experience around.
Machine Learning and A.I. Maximizes Value from Customer Support Centers

Success Stories — Crisis Management by Customer Support Centers

Nearly all customer support programs experience a surge in volume following bad publicity with customers threatening to terminate subscriptions, and they seem to be happening more often. Consumers prefer working with human representatives when it comes to stressful situations but unfortunately can be irrational. The customer support center’s job is to dissolve tension and salvage relationships. Robots do a very poor job at handling this and even human agents can fluster. The robot assistants are trained to help human agents follow the right protocols so even a less experienced staff can deliver best practice. Machine Learning and A.I. uses an analytical approach to provide customer service.

The Building Blocks of Intelligent Contact Centers

There is a symbiosis relationship between agents and technology to get the most out of the Machine Learning and A.I. algorithms. The genuine, complete dialogue records, along with client and account information, provide the data for model training. The customer service industry has been able to successfully implement this technology thanks to the large amounts of history that have been collected. Another reason for this progression revolves around the ability to perform topic modeling, which is used to identify intent and make associations to data in the knowledge base. There are many open-source language models that can be leveraged but domain knowledge is important to give context. A knowledge base is the collection of facts (e.g. customer interaction records, blogs, social media, FAQ, pricing, policies, and publications) about the business in the right context. The implementation of a robust system can be expensive and it begins with having rich data. The consolidation of information makes it easy to measure customer support center performance, which are the ability to help businesses raise CSAT and lower customer churn in order to generate (recurring) revenue while minimizing cost. Because of how fast the Machine Learning and A.I. fields are progressing, companies carefully manage their technology investments by weighing the benefits and cost. We’re starting to see industry leaders implement them but it’ll be some time before we see mass adoption.


Every interaction is an opportunity for businesses to create positive impressions with their customers. Machine Learning and A.I. are helping customer support centers in big ways but robot assistants are’t great at coming up with creative solutions themselves, that’s why human agents remain an integral part of the process. The result of human agents and technology working together means current problems are solved quicker, and future ones are eliminated, reducing customer support costs and increasing customer satisfaction in the process.

All Rights Reserved for Kevin C Lee

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.