Improving patient experience with contact center excellence
Delivering exceptional patient care is as important as ever in today’s healthcare landscape. Explore challenges & opportunities in how healthcare cont...
The Team at CallMiner
March 27, 2018
Voice continue to the most widely-utilized customer service channel by consumers, with 73% of consumers calling into the call center for customer service needs, according to Forrester. Other channels are gaining ground, however, with digital channels, such as chat and email, and web-based self-service becoming increasingly utilized by consumers.
Graphic via Sharpen
New technologies are providing consumers with more options for connecting with the companies they do business with, but technology advancements are also reshaping the way companies are meeting those needs. Once a pipe dream believed to be far off in the future, artificial intelligence (AI) is one innovation that’s transforming the customer service landscape. We’ve put together this guide to provide a comprehensive history of AI in the call center, from the advent of artificial intelligence as a whole to its first use in the call center and the potential for future disruption.
In this guide, we’ll discuss:
According to SAS, “Artificial intelligence (AI) makes it possible for machines to learn from experience, adjust to new inputs and perform human-like tasks.” Modern AI relies on technologies such as natural language processing and deep learning to train computers to perform certain tasks by processing and analyzing large volumes of data to identify patterns.
A 2017 article in Harvard Business Review (HBR) points out that machine learning (ML), which is the ability of a machine to continuously improve its performance without direct human input, is one of the most important AI technologies today. The biggest benefit of ML, according to HBR, is that humans often have difficulty describing precisely how a certain task is performed (such as recognizing another person’s face or making a strategic move in a game) – therefore, ML allows us to automate many tasks that we couldn’t previously automate. The authors predict that the effects of AI will be magnified over the next 10 years across every industry.
Graphic via Android Police
AI technologies are already outpacing human capabilities in some areas. Smartphone speech recognition, for example, is three times faster than human typing, according to a Stanford study conducted in 2016 by James Landay and colleagues. What’s more, the error rate has decreased from 8.5% to 4.9% – and these improvements have occurred in less than a year’s time. Another area of rapid improvement is in image recognition, such as the ability of Facebook to recognize a user’s friend’s faces in photos posted to the social network, or the ability of a smartphone app to identify a bird species based on a single photo with incredible accuracy. ML-based voice recognition is “now nearly equal to human performance” as well, HBR says – even in environments with noisy backgrounds.
Artificial intelligence and machine learning are also being leveraged by major enterprises for detecting fraud, money laundering, malware, insurance claims processing, and of course, customer support applications – which is the focus of this guide. However, before we dive into AI’s use in the call center, let’s take a brief walk through the history of artificial intelligence.
Artificial intelligence may be a buzzword today, the concept originated centuries ago among the classic philosophers of Greece who aimed to model human thought processes as a series of symbols. Its more recent history dates back decades to the 1940s with a concept referred to as connectionism, and in 1950, Alan Turing documented his ideas for testing a thinking machine. Turing’s test theory suggested that if a machine was able to communicate in full conversation via a teleprinter without any detectable differences from a human, the machine could be deemed “thinking.”
Two years later in 1952, the Hodgkin-Huxley model of the brain was introduced. The Hodgkin-Huxley model describes the brain as an electrical network formed by neurons. Individual neurons in the network function individually with an all-or-nothing on/off firing mechanism. The authors of the Hodgkin-Huxley paper, Alan Lloyd Hodgkin and Andrew Huxley, along with Sir John Carew Eccles, were jointly awarded the Nobel Prize in Physiology or Medicine for their discoveries in 1963.
The term “artificial intelligence” was first coined in 1956 as the topic for the Dartmouth Conference, which became the first conference devoted to the concept of AI. In the same year, the first running model of AI, Logic Theorist (LT), was demonstrated. Logic Theorist was written by Allen Newell, J.C. Shaw and Herbert Simon of the Carnegie Institute of Technology, today known as Carnegie Mellon University.
From 1956 to the 1990s, a number of AI-based models, games, and technologies were introduced. But in the 1990s, AI really started taking off. The 1990s marked a period of major advances across many facets of AI, from machine learning to intelligent tutoring, data mining, scheduling, virtual reality, and more. It was during the 90s that AI-based information extraction programs became widely used, such as web crawlers, which are now an integral technology for the World Wide Web.
In the 2000s and 2010s, AI hit the mainstream consumer market with a proliferation of smart toys – mini robots, interactive gadgets, and the like – as well as consumer products such as Amazon’s Alexa and Google Home. Alongside these innovations, AI has been making a major impact in the customer support industry, particularly in call centers.
Believe it or not, primitive forms of artificial intelligence were used in the call center decades ago, but only after decades of slowly progressing advancements in basic telecommunications technology. The switchboard was first patented in 1891 by Almon Brown Strowger and later acquired by Bell in 1916 – forty years after the invention of the telephone in 1876. Switchboards were the first automatic telephone switching system that allowed people to contact each other directly rather than relying on telephone switchboard operators.
Private Branch Exchange (PBX) and Private Automatic Branch Exchange (PABX) systems were the first technologies used to facilitate call forwarding to answering services for businesses (such as physicians) requiring 24/7 answering services, with some PBX and PABX systems dating back to the 1920s. But it was automatic call distributor (ACD) technology that made the concept of call centers possible, according to Call Centre Helper. While ACDs were quite basic compared to the technologies utilized by modern call centers, these systems make it possible to filter and route calls to the right agent at the right time – using an algorithm to determine the best agent to receive each call.
The first ACD systems are believed to have originated in the 1950s and were used by the major telephone companies. The Birmingham Press and Mail in the United Kingdom installed an ACD in 1965, which is the first documented use of this technology in the U.K.
By the 1970s, PABX systems were increasingly incorporating ACD, making large-scale call centers possible. In 1972, British Gas in Wales installed an ACD system capable of handling 20,000 calls per week, and this system is also believed to be one of the first to support multiple languages (in this case, Welsh and English).
Through the 1970s and 1980s, technological advancements continued to improve the call center, as call centers became increasingly important business functions. In fact, it was during this time that call center technology advancements made it possible for “companies to base their entire business model on telephone sales,” says Call Centre Helper.
Aspect Telecommunications, founded by Jim Carreker, was one such company paving the way for the future of call center operations by improving on early ACD technology – by improving the routing of calls from touch-tone phones, for instance, distinguishing different types of calls, and connecting calls to the appropriate team of specialized call center agents. These improvements enabled call centers to manage larger call volumes, driven by the introduction of toll-free phone numbers, and reduce call waiting times. Aspect would go on to become one of the world’s most prominent manufacturers of dedicated ACD technology.
In addition to PBX, PABX, and ACD technology, Motherboard identifies several other key technologies that made the call center possible:
Predictive dialers were first used in the 1980s as well. In these systems, computers call numbers from a database and distinguish human answers from busy signals, fax tones, and answering machines, routing answered calls to available agents. Predictive dialers use algorithms to determine the frequency at which numbers should be dialed, based on the average number of rings before calls are answered, the percentage of calls that are answered, and the average length of each phone call, as well as the number of live agents working at a given time. Predictive dialers automatically adjust their calling patterns continuously based on these factors with the goal of reducing the amount of time agents are sitting idle and achieving a desirable call abandonment rate.
While call center technology has advanced relatively rapidly since the earliest telecommunications systems, it was in the 1990s that technology innovations in the call center began to take the shape and form of those in widespread use today.
The rise of the internet drove the need for multi-channel customer support beginning in the 1990s, although “multi-channel” didn’t become a buzzword until the early 2000s. In addition to the introduction of SMS in 1992, the first commercial bots were developed in the mid-1990s. Chatbots arose out of a need to have someone present to interact with website visitors at all hours via chat interfaces. Bots are now a mainstream technology and are being used for everything from routing customer service requests to booking hotels, providing shopping assistance, and more.
AI’s use in the call center industry was, and continues to be, driven by the large volume of calls that call centers manage and the resulting time many consumers spend on hold. More than half of Americans (53%) spend 10 to 20 minutes on hold every week, while a majority (86%) are placed on hold every time they contact a company. Because these inconveniences have a negative impact on customer satisfaction, companies seek ways to more effectively route callers to the appropriate agent, reduce wait times, and enhance customer service. This is achieved with interactive voice response (IVR) and automatic call director (ACD) technologies, among others.
Today’s IVR solutions interact with callers to determine the caller’s need and route the call to the most appropriate agent. The result is that customers are connected with representatives who are best able to address their needs and spend less time being routed repeatedly to other departments and agents. After its initial use in the 1970s and the more efficient and cost-effective IVR solutions developed during the 1980s, IVR’s use in the call center first began to spread in the late 1990s and early 2000s.
However, the early iterations of IVR in call centers weren’t able to recognize conversations, only words or series of words that were pre-registered in the system. Today’s IVRs are increasingly equipped with speech recognition, allowing for the interpretation of a much broader set of words and phrases to more accurately ascertain a customer’s need and route callers to the most appropriate call center rep in less time.
While ACDs are intelligent systems, they’re not artificial intelligence in the truest sense – that is, they don’t think independently as a human would do. ACDs are conditional call routing solutions, based on if-then conditions, or rules pre-defined by the organization. Skills-based routing, circular routing, and most idle routing are a few examples of conditional call routing.
There is no single, universally agreed-upon line at which technology crosses over into artificial intelligence. According to Ramkumar Ravichandran, director, Analytics & A/B Testing at Visa, analytics is a part of the evolution that culminates with AI. Specifically, using predictive analytics to forecast likely future outcomes is often achieved by “using existing data to train predictive machine learning (ML) models,” as an article in ZDNet explains. “Whether being able to make predictions using machine learning constitutes AI, and whether having analytics in place is a prerequisite for this, are key questions to ask here.”
Predictive analytics is one modern capability that improves call center effectiveness, but it shouldn’t be confused with predictive dialing, a technology discussed previously that was first used in the 1980s. While predictive dialing is a valuable tool that can determine how many calls to make at the same time in order to get a live person on the phone in the shortest amount of time, thus reducing idle time for agents, this technology doesn’t actually do anything to improve relationships with customers. Predictive analytics, on the other hand, enables call centers to glean valuable insights in real-time, such as:
“Simply put, predictive analytics uses current and past information to create predictive models for the future,” explains John Ternieden in an article for Inside Sales. “It helps you anticipate potential outcomes and make more informed decisions, which ultimately helps you sell more.” Or, in the case of call centers offering customer support services, helps to pair callers with the agents most capable of addressing their concerns, arms call center reps with pertinent information about the customer’s history and enables them to craft tailored messages perfectly in tune with the customer’s problems, preferences, and needs, and even aids in identifying customers who are more likely to be satisfied with call outcomes.
Speech analytics is another newer technology increasingly utilized in the call center. Also known as voice analytics, this technology was first used in enterprises such as call centers in the early 2000s for commercial purposes, and the market is expected to reach $1.33 billion by 2019. It goes beyond recognition, interpreting not just the words a caller speaks but also the manner in which those words are spoken. Speech analytics detects factors such as tone, sentiment, vocabulary, silent pauses, and even the caller’s age, analyzing these factors to route callers to the ideal agent based on agents’ success rates, specialized knowledge and strengths, as well as the customer’s personality and other behavioral characteristics.
In addition to analytics, the modern use of AI is closely interwoven with concepts such as machine learning (ML), data mining, big data, and automation. Combining AI with technologies such as predictive analytics can result in a more powerful, more scalable, and more efficient application of data. Businesses are taking note, as well: According to Karl Flinders in a January 2017 article for ComputerWeekly.com, 64% of businesses say that their future growth hinges on large-scale AI adoption. Those that have adopted AI technologies say they expect their revenues to increase by 39% by 2020, while costs are expected to drop by 34% by the same year, although concerns such as data security, job security, and pay rates are prevalent.
In the call center, AI is increasingly utilized, but technology isn’t expected to make human call center agents obsolete anytime soon. As ValueWalk explains, chatbots aren’t precise enough to adequately address customer or user needs in some cases (Facebook’s Messenger bots have failed 70% of all user requests, for instance), and customers still desire human-to-human interaction. Customer satisfaction is based heavily on how the customer feels they’ve been treated, and today’s consumers demand personalized, individualized, relevant experiences with the companies they do business with.
AI is useful for enhancing the overall experience, but shouldn’t replace the personal interactions that drive brand reputation and customer satisfaction. For instance, AI technologies such as chatbots can engage prospects with coupons or savings opportunities, allowing human sales reps to provide that all-important personal touch to close the deal. Likewise, AI can arm call center agents with robust historical data and insights about a customer, empowering agents to deliver meaningful cross-selling and up-selling opportunities, which translates directly to a healthier bottom line.
What’s more, AI solutions can predict the best times to engage prospects or customers with information that’s relevant to them in the moment, reducing the frustration consumers often experience from being bombarded with irrelevant messages at inconvenient times. In short, AI will make call centers more efficient, augment call center agents’ knowledge, and improve experiences for customers, while simultaneously satisfying customers’ conflicting demands for automated service and personal interactions. The key for call centers today and in the coming years is to find the right balance.