The Changing Landscape
In 2013, economists Carl Benedikt Frey and Michael Osborne of Oxford University found that a whopping 47% of jobs were at risk of displacement due to the coming Artificial Intelligence (AI) revolution in the U.S. labor market. Telemarketing came in at #1 as the most likely job to be automated (99% probability). Data Entry Keyers were a close second (also 99% probability). Not far behind were Dispatchers, Receptionists, and Information Clerks (96% probability).
This is Part 1 of a series of articles that focus on the incredible impact of AI on the contact center industry. Part 1 deals with the causes behind this impact, gives specifics on how this impact is taking place, and offers a rationale for why contact centers must take note and prepare for the revolution. Part 2 of this series will focus on the Total Cost of Ownership (TCO) of the New Contact Center, and make the case for AI investment using a Return On Investment (ROI) analysis.
“33% of contact centers plan to invest in robotics and process automation in the next two years”
From a 2017 Deloitte survey of over 450 global contact centers:
- Social media will account for 9% of total contact center contacts in 2019
- 33% of contact centers plan to invest in robotics and process automation in the next two years
- 56% of technology, media, and telecom (TMT) companies are planning to invest in AI
- 85% of organizations anticipate contacts will become more complex in the next two years
- Phone is expected to only account for 47% of contacts in 2019 falling from 64% in 2017
Better, Faster, Cheaper
“More than 1 million jobs are at risk in four countries alone: the United States, Poland, India, and the Philippines.”
The ratio of human-powered work to automation-powered work can – and will – shrink. A report by management consulting firm A.T. Kearney in 2016 concludes that, a software robot costs one-third as much as an offshore employee and one-fifth as much as onshore staff. A.T. Kearney estimates that robotic process automation can yield cost savings of 25%-50% for many business processes. Indeed, A.T. Kearney reports in 2017 that “More than 1 million jobs are at risk in four countries alone: the United States, Poland, India, and the Philippines.” The report goes on to state the following:
“In fact, automation creates new, more highly skilled and more highly paid jobs that are required to manage mature technology, bringing higher salaries to workers. These new jobs, however, do not always go to those who lost their positions or even stay in the same country, as some new jobs are created in closer proximity to existing operations. The overall impact of automation on a country’s labor force, then, is far from straightforward.”
The Transition to the New Contact Center
As call centers automate and digitize, utilizing technologies such as AI, they will need to develop skills in managing automated workflows, and training approaches for AI, in addition to servicing increasingly complex customer needs. Success at developing these new skills and seamlessly integrating the technology into call center processes is paramount.
“Businesses believe that customer experience drives customer choice”
The changes that are coming to contact centers are not just due to the level of automation that they can employ, they are demanded by their customer base for the following reasons:
- Businesses believe that customer experience drives customer choice – the abandonment rate for contact centers at contract renewal has been 50% over the last two years – contact centers are transitioning from cost centers to profit centers
- Contact centers are being asked to step up, to transition to a broad mix of channels, match self-service technologies to simple interactions, and hire and equip contact teams to handle increasingly complex interactions
Another 2017 report has quantified the changes in contact center job profiles on a regional and skill level basis. It shows that job losses are primarily in the low skill level, ranging from 25% in the Philippines to 61% in the US. Jobs in the medium skill level marginally increase across all regions, while jobs at the high skill level show the greatest growth, ranging from: 45% in the US, to 59% in each of the UK and India, to 100% growth in the Philippines.
Can AI Agents Completely Replace Human Agents?
Evidence exists that, where customer interaction is concerned, human customers often prefer to deal with other humans rather than with computers, particularly where advice is being sought.
“AI technology is not yet ready to take over the job of a human being entirely, in areas that require human contact”
A 2016 Accenture Global Consumer Pulse Survey indicated that 83% of US consumers prefer dealing with human beings. Part of this preference is likely linked to the sub-standard performance of human-computer interfaces in the past. On the flip side, human operators can only deal with so much information. Today, there are literally gigabytes and sometimes terabytes of data available that a human operator couldn’t possibly access in a meaningful way, but that could potentially contribute to an improved customer experience (CX). How do we marry these two apparently conflicting points to arrive at an improved CX?
AI technology is not yet ready to take over the job of a human being entirely, in areas that require human contact, for two reasons:
- AI cannot pass a well-designedTuring Test – Even though AI can converse through text e.g. chat, SMS, and through voice e.g. NLP over VOIP, AI cannot “reason” sufficiently to convince a human being that it is, in fact, another human being – the essence of the Turing Test. An example of such a conversation would exhibit natural language, reason, knowledge, and learning capability. Turing also envisioned emotional and aesthetic intelligence in some of his example conversations. Note: The Turing Test has undergone a good deal of discussion since published in 1950, but the original intent of the test demonstrating that a computer could imitate the responses of a human has remained unchanged.
- Humans still prefer to deal with humans – Once people discover that a digitized solution can provide a superior CX, some of this bias will most likely disappear. An example would be ATM machines, and online banking. Since the first cash dispenser was introduced in 1969, getting money when you needed it 24/7 has become a common convenience and an essential way we spend our money. According to a 2015 US Federal Reserve report, 75% of consumers in the United States use an ATM as part of their everyday banking activities, while mobile and online banking (74%) usage is significantly increasing. Clearly, convenience can override human interaction preference if the service is improved overall. Online travel bookings are another example of improved service overriding human interaction preferences. Online gross travel bookings revenue worldwide moved from $340B USD in 2011 to $567B USD in 2017. A 2017 report on ATM statistics states that brick and mortar stores that install an ATM experience a 20% increase in sales, while ATM users on average spend 23% more than non-ATM users. Clearly, improved CX can also influence top-line growth.
“each of these capabilities… when combined…become a powerful tool…drawing out knowledge that would otherwise never have been available, and applying that knowledge in real-time during a call”
While we can’t entirely replace human agents with AI agents for all functions, we can introduce an AI Virtual Assistant that can participate in the interaction in the background, assisting the human operator while not being apparent to the customer. To that end, our AI Virtual Assistant is going to be good at 4 types of processing:
- Big Data – finding patterns in large amounts of data that exceed the capacity of traditional databases
- Natural Language Processing (NLP) – parsing language as spoken and written by humans (such as in Amazon’s Alexa), including in multiple languages
- Machine Learning – training generic algorithms e.g. neural nets on Big Data datasets using NLP, as opposed to human based programming of rule sets
- Internet of Things (IoT) – a source of Big Data, IoT will eventually provide real time input to our AI Virtual Assistant.
Taken alone, each of these capabilities could prove of some value. When combined, however, they become a powerful tool that allows us to take resources that previously were of little value – such as hours of call recordings – drawing out knowledge that would otherwise never have been available, and applying that knowledge in real-time during a call. Let’s use an example narrative to better illustrate how such an AI Virtual Assistant might aid a human operator.
The New Contact Center
Imagine if you will, a contact center that provides phone and dispatch support for vehicle road side assistance for association members. While the association members are in America, the contact center is in South East Asia. It’s late at night in Boston, in the dead of winter, when the phone rings in the contact center and a human agent moves to answer the call.
“we can expect [these technologies] to become common place over the next 5 years”
Anticipating Customer Needs
Even before the phone rang and while the call still sits in the queue, the contact center’s AI Virtual Assistant judges it to be urgent. In a few milliseconds, it made that determination by acting on the context of the call:
- The caller-ID was associated with the callers account.
- She has been a customer for several years but hasn’t called the rescue line once.
- Other customers with a similar profile generally call only when they really need help.
- The weather in the customer’s home city is cold enough that hypothermia is a risk if she is stuck without heat for too long.
Based on that context information, the AI Virtual Assistant moves the customer call to the top of the queue to be answered next – and in seconds, the phone of the human agent rings and he moves to answer the call.
All of the above occurred before the phone even rang in the contact center. What just happened here? Let’s take it point by point:
- Linking the caller-ID to the callers account can be done by the AI Virtual Assistant because it sees all the calls in the queue, and can act in milliseconds while humans are processing other calls. It would take some time for a human agent to do the same, perhaps a minute or more for each call in the queue. Since the AI assistant operates a million times faster, the overhead is negligible.
- Knowledge that the customer has never made a rescue call in several years, can be obtained in a number of ways once the link to the account is established.
- Profiling of customer call patterns is based on applying machine learning to Big Data, something that is not done today. This profile information can be associated with individual accounts allowing insights regarding customer behavior.
- Weather is just one of the feeds to the AI Virtual Assistant that can be easily accommodated and yield valuable information relevant to the local time and location of the caller.
The human agent answers the call. As the customer details the problem, and describes her location…
- While the human agent elicits information from the customer, the AI Virtual Assistant, which has been listening to the call in the background using NLP to extract keywords, pulls up a map of the location including the position of various roadside assistance trucks nearby.
- The customer asks for a tow, and the agent responds that they will send one right away. The AI Virtual Assistant sends an automated request directly to the nearest tow truck with the customer location information (much like Uber). The AI Virtual Assistant receives an estimated time of arrival from the tow truck software which is displayed on the contact center screen. The human agent informs the customer that the tow truck will arrive in 7 minutes, and concludes the call.
- All information relevant to the call and its outcome are recorded for posterity by the AI Virtual Assistant, while the human agent is immediately available for the next call
The above description is not all that far fetched, and we can expect scenarios like this to become common place over the next 5 years. In 2016, Vonage partnered with IBM Watson to work on a number of AI integration solutions. Both companies have packaged their products in order to provide an Application Programming Interface (API) into their respective functionalities. Vonage acquired Nexmo in 2016 to provide such an API into their own products, while IBM created Intu, a systems-agnostic version of Watson that allows developers to embed artificial intelligence (AI) functions into devices, robots and applications. This is one example of the partnerships that will evolve to take contact centers to a new level of excellence.
Not all AI improvements will require such a level of sophistication. Other automation opportunities exist when interacting with customers such as account profile update, and Frequently Asked Questions (FAQs) that might be entirely handled by an AI Virtual Assistant through a chat, text, or even a voice based interface. If the AI agent detects that a human agent is required it can route the call appropriately.
Stay tuned for next article in this series – Preparing Contact Centers for the Advent of AI – Part 2: Estimating ROI and Benefits. In Part 2, we propose to cover:
- Improved agent handle time
- Infrastructure cost avoidance savings
- Reduced head count due to improved agent utilization or staffing automation
- Increased net revenue through improved eCommerce conversions
- Increased net revenue from improved voice conversion rates: upsell and cross-sell
We’ll cap that discussion off with the presentation of an ROI calculator. We’ll see you then!
Glenn Reid is the CEO of RJB Technology Inc., a Canadian firm with Branch Offices in Makati, Philippines. Contact us today for more information about our company, or to discuss your custom software development needs.