Will AI Own Customer Service? Why Handing Over the Reins Isn’t Always Easy
With the increased onset of digital channels for marketing and the influx of data that follows, every marketer is seeking a consistent brand experience across channels, and the AI technology we have at our fingertips can play a significant role. But, other than some point applications with chatbots and the like, businesses are hesitant to turn the customer service reins over to AI. But why?
Simply put: fear and confusion are hindering marketing practitioners from utilizing AI to its full potential in the marketing realm. Too often, we use buzzwords that define the extreme, keeping professionals from understanding the root of a particular technology; in this case, the buzz-term is “Artificial Intelligence.”
Most people don’t always think positively about artificial intelligence. Their minds immediately jump to HAL or SkyNet. If you’re not familiar, HAL (short for Heuristically programmed Algorithmic computer) is the “bad guy” in Arthur C. Clark’s Space Odyssey series, where the computer tries to kill off the crew “he” is designed to protect in order to preserve his own existence. Similarly, SkyNet is the sentient neural computer network from the Terminator movies that achieves “singularity” — the moment when AI becomes smarter than its designers and is truly “aware” of itself— and essentially starts to take over as a means of self-preservation.
With this as the primary mental image of artificial intelligence, it’s no wonder that fear starts creeping in when someone mentions applying AI to marketing and customer care. Who in their right mind would want to turn over customer care to HAL or SkyNet?
AI’s Not-so-Scary Reality
If you look at the sub-categories of artificial intelligence however, there is much more comfort to be found. AI is often described as being capable of speech and facial recognition, natural language processing, automated reasoning and even playing games. When we examine these sub-topics alone, HAL doesn’t seem so scary after all. In fact, he sounds like a combination of Siri and Google Photos. Sure, it’s a little disconcerting that Google Photos can recognize your face out of a sea of other images, but you hardly fear that it is about to achieve singularity and wipe out the human population to preserve its own existence by doing so.
When viewed as sub-capabilities, artificial intelligence is already being used successfully in all manners of customer interaction and care. When a customer dials into a call center and is met with the message, “In a few words, please tell us why you’re calling,” they don’t immediately think that the world is coming to an end. (Of course, we’ve all been trapped in automated voice response purgatory at least once, screaming “representative” at the top of our lungs — but those instances are fewer and farther between.)
There are even examples of speech recognition in customer care for healthcare and insurance where the callers’ speech is used to intelligently route them to the best-qualified agents to address their specific needs; or, when used by a service agency to schedule a pick-up or set up an appointment.
Oftentimes, we prefer the terms machine learning or deep learning as compared to artificial intelligence — myself included. Neither of those terms connotes the extreme of a truly sentient being that’s capable of going completely off the rails; but rather using cognitive technologies to improve and automate certain areas of business decisioning, including customer care and many broader aspects of customer interaction management.
Spotify and Netflix are already using machine learning to help users expand their playlist — and that’s not scary at all. Along the same vein, Amazon and Overstock.com can make great product recommendations based on my interests, and that’s not scary either.
Given these examples of non-threatening, current uses of AI, we see that the primary hurdle is actually organizational in nature. Simply put, the humans involved either don’t trust the technology or don’t want to give up control.
Handing Over the Reins
Product recommendation engines, product affinity models, collaborative filters, churn/retention models and next-best-offer calculators all use embedded machine learning capabilities to make their decisions.
Unfortunately, most of these providers “bury” their machine learning as esoteric algorithms deep within their code, making it nearly impossible for business leaders to know how its making decisions, much less proving specifically where AI is providing value. It’s really hard to trust the black box in the corner that nobody can quite explain. It’s almost like a math teacher thinking a student successfully guessed the correct answer without showing any of the work.
The most successful AI will provide reports, insights, and feedback on exactly what its doing. AI technology needs to “show its work” so the humans involved can better trust the decisions that it is making. Additionally, the more decisions and outcomes AI makes and sees, the more it can determine which attributes have the highest predictive influence. Knowing what data and attributes AI uses most effectively to make its decisions can help marketing teams provide even better, fresher data into that decisioning environment moving forward.
Giving Credit Where Credit is Due
All this aside, the biggest issue that sentient humans have with AI is often all about politics and control. If I’m the product manager for Product A, for example, and I want my product offer extended to everyone that I’ve targeted as likely consumers, I don’t want a piece of technology deciding that Product A will not be offered to a specific customer for reasons unbeknownst to me. I want control.
Politically and organizationally, it is often very difficult for product, channel and brand managers to let technology make decisions, especially as they pertain to customer segmentation. Much like the trust dynamic in the previous paragraph, we need AI technology to provide insight to the business that actually proves — to all managers involved — the lift it provides, the reduction in support cost achieved, and/or the overall improvement it makes in terms of return on marketing investment. This provides validation to the business around the strategic impact of AI, while also allowing the management team to “take credit” for business improvements.
While part of the issue with AI continues to be fear and confusion — which is relatively easy to remedy — the more difficult hurdle to overcome is the perception that using AI minimizes the role of management decisions. If AI is deployed correctly, it will have the ability to “show its work” and prove its value to the business in order to justify continued deployment.
AI is here to stay and will continue to become more prevalent. The biggest need in the industry today is one of education to eliminate the misunderstanding of exactly what AI is: AI is your friend and it’s here to help.
About the author: John Timmerman is a global industry evangelist for Teradata where he has worked for the last 23 years. John has seen all sides of the Teradata enterprise, including most industries and geographies, through his work in sales, business development, sales support, product management, and marketing. For the last 12 years, his focus has been in the areas of CRM, Customer Interaction Management and Inbound Marketing.