Exploring AI's Impact on Call Center OperationsExploring AI's Impact on Call Center Operations

The future of artificial intelligence (AI) is the subject of a lot of buzz and speculation about if and how it will take over for human tasks. Recently I heard a great comment, suggesting that if a human can do it so can AI, but if it hasn't been done in AI before it may take some serious research and development.

This reminds me of the relationship and differences between Henry Ford and Thomas Edison. Ford's genius was in automation and the assembly line. He introduced the specificity of each person on the line performing a single task. The workers became skilled at that one task due to repetition, but they quickly became bored with the monotony. Edison, on the other hand, embraced an "innovation through experimentation and failure" model in which experimentation and trying different things came together to make breakthroughs.

The application of AI involves recognizing that people don't want or need to do the mind-numbing tasks for much longer -- computers can handle these tasks and have already started to in many industries. But in working with AI or training AI to take on advanced tasks, we need some creativity and the freedom to learn through trial and error if we're to develop a working system of innovation.

In the customer service realm, AI will continue to transform how customers interact with both human and artificial agents. But will the result of this transformation mean the elimination of most customer service jobs? AI will fundamentally change such roles as it takes over assembly-line type tasks, but the need for human agents to answer questions and data scientists working behind the scenes to innovate remains strong. According to Gartner research, by 2020 AI will be the driver that creates more jobs than it eliminates, a dynamic that's likely to be true within customer service.

Exploring AI within the Call Center
Advanced AI tools are perfectly suited for extracting mundane tasks from call center agents. The industry leaders in call center solutions are using AI to find patterns in data and then present that to the agents and company in an automated way. The result is instant context, and the ability to send some calls to automatic responders and others to human agents, as appropriate. Call centers can then task their employees with more enriching queries and jobs, and develop their critical thinking and communication skills.

Consider the McDonald's drive-thru window for an example of such natural language AI at work. The company utilizes voice recognition AI to transfer a customer's order to the payment system and food prep line. It's more accurate than the notoriously scratchy drive-thru speakers and helps remove a layer of mundane tasks through automation while not replacing the need for workers to fill the orders. And of course automating the order-taking helps speed up the drive-thru process with on-screen order confirmation and freeing the employees from being glued to the register to enter orders.

Despite the promise of AI to add context and instant data to any interaction, it does have limitations. False results are a prominent limiting factor of AI's current usage within service centers. Consider a customer who orders a premium sports package year after year. The customer calls into the center knowing that he'll order the package at full price if it comes down to it, but he's still looking for a discount and threatens cancellation in the hopes of receiving a discount offer. If the customer speaks to an agent who is aware of this common practice, the agent would know not to offer a minimum or no discount at all. If an AI engine trained to pick up on any mention of cancellation but without the machine learning capability to understand context handles the call, it might erroneously offer a steep discount to ensure against cancellation.

Continue to Page 2: Is Replacing Humans on the Horizon?

Is Replacing Humans on the Horizon?
There is a great quote by computer science pioneer Alan Perlis: "A year spent in artificial intelligence is enough to make one believe in God." That's not to say that AI is omnipotent. It's great at finding patterns, but it still needs the human in the loop to create the framework. AI doesn't work by simply shoveling all sorts of data into a computer oven that magically bakes relevant insights. The human mind still needs to structure and wrangle the data in order to give it relevance, because AI is still very deterministic and requires a human neural net. Data scientists are needed to train the AI to understand the right context so the programs can complement, instead of replace, human intuition and intelligence.

Today, whenever you have AI between the customer and a human, you're likely to have natural language errors. These types of errors can be cute when they happen with Alexa or Siri, but the customer won't be amused if the AI is making cheeky comments about medical data or financial records. The neural net driving current AI platforms currently can't keep up with all of the routes conversations can take. As a simplistic example, if the customer is on the phone and has to yell an aside to a child, "Timmy, you're in timeout," the AI might deduce the customer's device/program is "timing out," while the human of course will know Timmy is a misbehaving kid. In another scenario, if a customer is calling to complain about a missed delivery of a critical item, the AI machine voice may have difficulty de-escalating rising emotions. Currently a human touch is needed to exhibit the right amount of empathy.

Humans are still very much needed for customer service interactions that are more complex or involve a multi-layered explanation that takes into account many data points across disparate topics. Customers certainly aren't ready to talk to a robot voice when discussing sensitive or emotional topics or other scenarios where health hangs in the balance. A lot of calls cover many different tangents, and we just aren't there yet in terms of an AI that can think around all of the boxes and paths. Errors still happen, although the industry is pushing more testing to get to a state where information is propagated by the AI automatically with few or any errors at all.

A likely future for AI within the context of customer service is a three-way conversation between the customer, a human agent, and an AI-powered assistant. The agent can ask data-related questions of the AI bot during a call, which can then quickly pull up needed information. This approach allows the agent to answer questions immediately and provides a chance to build the customer's comfort with the AI robot. Perhaps next time that customer calls he'll have the option to just speak to the AI assistant directly. However, even in this scenario the human is still in the loop and will need to develop a partnership relationship with the chat assistant. Call center managers and staff should embrace AI as a tool to improve their job performance and decrease mundane tasks. And ultimately, it's a tool that can help delight the customer's expectations, and any agent that can accomplish this goal has a nice amount of job security.