Automation: The Contact Center’s Grocer

In times of crisis, we often find heroes in the unlikeliest of places. The current COVID-19, or coronavirus, pandemic is no exception.

The obvious heroes of the current crisis are, of course, the health care workers and first responders who are on the front lines caring for the sick and maintaining a brave demeanor in the face of a health emergency the likes of which the world hasn’t seen since the influenza pandemic of 1918. They put themselves in harm’s way to tend to and care for the sick, and so the rest of us might be safe from the COVID virus.

But what about the less obvious workers who also risk illness or worse as they perform their jobs daily? These are the people who work so the rest of us have the food and other supplies we need as we are safely quarantined at home. Even with social distancing, these workers often find themselves in close quarters with the public for hours at a time, and this goes on day after day. I’m speaking, of course, about your local grocer and all the people who are working in grocery stores through this health crisis.

I think it’s safe to say that the public in general has always been grateful for, and have admired, health care workers and first responders for what they do. But when was the last time you walked into a grocery store and thought to yourself, “I’m so glad these workers are here stocking shelves and bagging my groceries so my family will eat tonight?” I daresay grocery store workers are a pretty much overlooked bunch, but they perform a critical function that all of us benefit from, especially in times like these. And sometimes, for a few brief moments, they are recognized for their efforts. I bet when you went to the grocery store last month and finally saw toilet paper and bleach on the shelves, the value of those grocery store workers went up a notch or two in your book. The work that the grocery store employees are doing is exceptionally important during these difficult times.

Coincidentally, there is a similar phenomenon occurring in the contact center industry. Automation is a technology we tend to take for granted in the contact center, and not surprisingly so. Automation is defined as technology that enables a process or procedure to be executed with little or no human assistance, and in the contact center automation has been an integral cog in the customer service engine since the beginning. The first automatic call distributor (ACD), which automatically routed a customer call to the first available agent, was delivered to Continental Airlines in 1973.

In the past few years, automation in the contact center has seen a resurgence with applications being automated on both the agent side and the customer side. Commonly referred to bots, digital agents or intelligent virtual agents in the industry, interest in automated processes has been steadily growing over the past few years.

Table 1 illustrates the take rate of artificial intelligence (AI)-driven virtual agents in the North American contact center industry. The data is from the 2017 through 2019 surveys of customer service professionals that Saddletree Research conducted in conjunction with the not-for-profit National Association of Call Centers (NACC).

The chart illustrates the steady if unspectacular growth of AI-enabled virtual agents with over 16% industry penetration by the end of 2019. Revenues were in the neighborhood of about $14 million, which is not bad for a relatively new market segment. I believe, however, that this growth trajectory is about to radically change.

The COVID-19 pandemic has put a stark spotlight on AI-enabled virtual agents, whether they are saving the day for some, or could have saved the day for others. In the case of filing for unemployment, for example, virtual agents have definitely saved the day for so many.

DoNotPay is a small company in San Francisco that bills itself as “The World’s First Robot Lawyer.” The DoNotPay.com application initially provided assistance for routine legal and similar problems such as appealing a parking ticket or filing a lawsuit in small claims court, and assistance is provided by intelligent bots or agents. The seven-person company, however, saw an unfulfilled need created by the current pandemic and deployed its bots to help.

With unemployment rates in the U.S. closing in on numbers that haven’t been seen since the Great Depression, filing a claim for unemployment benefits has become a nightmare. Call centers are overwhelmed with more calls than they can handle, and frustrated callers are being met by a recording that simply tells them there are too many calls waiting and to call back another time.

DoNotPay has deployed its intelligent bots to gather information from people wishing to file for unemployment and completes an application for them online. The filer then prints the application and sends it in via snail mail. Incredibly, these applications are being processed faster than applications completed online or via the phone. The reason? The government systems used to process unemployment claims date back to the 1960s! As a result, 1960s-style written applications received via snail mail are more quickly and easily processed than online or telephone-based applications.

What sets automated intelligent bots apart from other AI-driven solutions is that it is a solution within the reach of virtually any contact center, regardless of size or resources. It is not a solution that requires expensive integrations or complex installations.

To better understand the evolution of intelligent virtual agents, I enlisted the help of Chris Crosby, CEO of Xaqt in Chicago. I’ve known Chris since 2005, when he was the founder and CEO of Latigent, which was acquired by Cisco in 2007.

Stockford: What is it about intelligent virtual agents that has made it such a democratic solution, available to contact centers of all sizes?

Crosby: When virtual agents first arrived on the scene, they tended to be expensive chatbots on company websites that often took months to develop. For the most part, they had little functionality and failed to deliver on their hype and promise. However, recent advancements in deep learning along with inexpensive cloud resources and big data combined with the mainstreaming of cloud-based communications platforms (CPaaS) have provided all of the ingredients to create unified agents that can deliver exceptional customer experiences across any channel including, natural language voice and SMS.

Also, most chatbots were developed as a one-time project and were not adaptable. Meaning someone would build the bot and it would remain a static chunk of code that was outdated the moment it went live. Whereas virtual agents that are delivered in more of a “Conversations-as-a-Service” model where the vendor provides the ongoing performance monitoring, tuning, training and enhancements as part of a usage-based pricing model reduces the overall expense and ensures that even incremental improvements compound over time yielding optimal results.

Stockford: What is the state-of-the-art today in terms of AI-enabled virtual agents?

Crosby: I think it’s two-fold. First, for AI to be effective, it requires a lot of data and supervised training by data scientists. That’s no easy task. The ability to train linguistic models at scale and deploy those to multiple call centers in similar verticals or with common call types creates economies of scale. Second, is the ability to combine conversational AI with robotic process automation to create virtual assistants that can both comprehend what the customer is asking of it and is also capable of executing a workflow, such as updating a customer record in the CRM or routing the customer to the best agent.

On one end of the market, we’re still seeing companies building bots with “Do-it-Yourself” tools from the cloud providers. This inevitably leads to frustration and poor customer experiences as these tools have limited capabilities and lack the massive amounts of tailored training data required to achieve the language recognition accuracy you need to ensure a quality customer experience. In this scenario, you’re basically building your own bot in a silo with no context and then you’re on the hook to constantly babysit the thing to make sure it works okay. (Oh, and you have to build all of the performance monitoring framework behind it, too.)

On the other hand, savvy companies can now personalize each customer interaction by infusing a virtual assistant with machine learning at every point in the interaction. Now a bot can do things like make hyper-targeted upsell or cross-sell suggestions based on data from a CRM just like a live agent might. With the right training, a virtual agent gets smarter with each interaction. Companies that go this route will have an advantage and be able to use their customer experience as a differentiator.

Stockford: I think the industry has learned some valuable lessons as a result of having to scramble in order to meet CDC guidelines while maintaining customer service levels. What do you think is next for the industry as we start to emerge from the darkness and get to the other side of the pandemic?

Crosby: I see three phases unfolding in the coming months.

The first stage has been triage, where companies in crisis looked to virtual assistants and automation to remediate a deluge of call volume for things like reservation cancellations or to handle FAQs for inquires related to COVID-19. Most of these bots are simplistic in nature but served their purpose.

The second stage will take shape as companies begin to reassess the role that virtual agents and automation can play in both their customer experience and their call center operations. This will require massive education in the market as to the true capabilities that the technology can offer. It will also require a normalization of KPIs and a shift from bots being built by IT to call center operations viewing virtual assistants as more of an extension of their team.

As example, the key measure of a bot’s success today tends to be “containment rate,” or the percentage of calls that were handled by the bot or deflected from live agents. This is similar to how IVR performance has traditionally been measured. While that’s important, it’s more critical to be able to score the virtual assistant’s performance more akin to that of a live agent in terms of customer experience and operational performance. I think you’ll see call center mangers keenly focused on things like bot handle times, first-call resolution and even sales conversion rates. At this point, virtual assistants will move from being a stopgap to save costs to being a strategic asset that customers want to interact with.

In the third stage, omnichannel virtual agents become the standard entry point for customers into an organization. One common misconception about virtual assistants is that they are “all or nothing,” meaning that the goal should be to fully automate a customer interaction. In reality, implementing a natural language IVR or even offering automated SMS interactions can afford many benefits, such as authenticating a customer or capturing key information before routing to a live agent. Not only will this improve customer experience, but for most companies, reducing agent AHT by 20-30 seconds is meaningful. Additionally, for calls that can’t be fully automated, the ability to enable contextual-based routing will be transformational.

Adoption is underpinned by the fact that these new AI omnichannel communication pathways can be quickly integrated into any contact center environment, even on-premise ACDs without SIP trunks.

Stockford: We’re hearing stories of how intelligent virtual agents have saved the day for many people during the pandemic, as well as hearing the “Coulda-woulda-shoulda—if only I had automated,” stories. I’ve revised my forecast for market segment growth over the next five years to show a pretty impressive upward trend, but in terms of applications, what’s next for intelligent virtual agents?

Crosby: Advancements in AI and robotic process automation (RPA) will enable virtual assistants to do more things. Contact centers will start to move past the notion that virtual assistants merely automate interactions with customers, and we’ll see expanded use cases. For example, the opportunity to provide agents with personalized assistants that provide access to real-time information and guidance is immense. This could range from simplifying a complex knowledge base to making real-time recommendations to customers based on what’s being said in the conversation. Purpose-built bots will begin to automate processes such as alerting supervisors when agents are on a difficult or long call using real-time sentiment analysis combined with ACD data, or even automating outreach to agents when additional staffing is needed. Really, many interactions that employees have today with their supervisors or managers can be automated using virtual assistants. This will free up resources for higher value activities.

Stockford: Clearly, automation and AI-enabled virtual agents will continue to grow in importance to the industry, which led me to revise our market forecast to show a compound annual growth rate (CAGR) of over 64% between 2020 and 2024 (see Figure 1). Impressive numbers for an impressive contact center solution!