I’m not known for being quick on the uptake by those closest to me. A couple prime examples include long division and driving a stick-shift. When my only option was a manual transmission, my parents just sort of cut me loose on the streets of Rancho Cucamonga, California to learn to drive (Yes, I already had my driver’s license but on an automatic transmission).
What followed was a couple solid weeks of stalling that red 1987 Toyota Corolla at every intersection with my two little brothers laughing at me from the back seat. Perhaps out of pity or because I had jostled a headlight loose, my dad finally took me out for a lesson and helped me understand the balance between the clutch and the accelerator. Within fifteen minutes, I knew how to operate a manual transmission without stalling. Without that short lesson, there’s no telling if I would have ever figured it out on my own.
So keep that in mind as I talk about technology for customer experience. I’m slow to grasp new concepts but once I do grasp them I generally understand them very well. That being said, I thought it might be helpful to share what I understand thus far about a couple concepts that are often mentioned in the CX technology space.
Do terms like NLP and Machine Learning mean anything to you? It’s OK if they don’t but chances are you’ve heard a well-meaning salesperson try to sell you on the latest greatest CX tool and they’ve dropped these terms like they’re hot. For the sake of clarity, let’s break these terms down a bit.
Natural Language Processing (NLP)
NLP enables computers to process large amounts of text in any language and understand what’s being said. This is exciting because we communicate with our customers via text whether it’s through inbound support messages, transcribed phone calls, or voice of customer survey responses. The big challenge is parsing through the data to get actionable insight and NLP is foundational to this work.
Let’s look at an example where we see NLP at work in the CX. Imagine you’re on a company’s website and are searching through their knowledge base for an answer to a question before contacting customer service. Traditional searching based on keywords yields results but with far less accuracy. NLP on the other hand understands what you’re searching for and the context around it to provide a more accurate recommendation. This is basically what all of the buzz around IBM Watson has been since beating Ken Jennings on Jeopardy nearly a decade ago.
Now take into consideration chatbots or any sort of automated response to a customer. This is something that’s received a lot of buzz in CX communities over the past few years. The ability to respond to customer inquiries is all predicated on understanding accurately what the customer is saying or asking. This doesn’t happen without NLP.
The second important concept in this mix is Machine Learning. This is the process of training or conditioning machines to respond accurately. The best way to do this is by feeding it data, lots of data.
Here’s an example from the text analytics world. Let’s assume you want to be able to predict if your customers are going to be satisfied with a particular response from customer service. Even better, what if you could give your agents not only the correct answers but the best way to answer ahead of time? You’d want to first build a machine learning model where you feed it thousands of customer interactions and tie those interactions back to success metrics like Net Promoters Score, Customer Satisfaction, or closed sales. Pretty soon, the machine will be able to understand the patterns that lead to success or lack thereof.
Going back to chatbots, you can totally train a bot to automatically respond to customers, and chances are they can do it with more consistency and accuracy than a human. But this first requires a ton of data to build of the confidence level that the bot is accurately answering the question posed by the customer.
The next time you search a knowledge base or interact with a chatbot and it asks you if the answer it provided accurately addressed your question, consider that you might actually be training a machine and helping it gain confidence. Contact center agents might also be assisting in this work every time they apply a canned response or macro to a ticket or a chat.
I recently wrote an article for ICMI with technology upgrades for contact centers to consider in 2019. In that article I cautioned us to not become too enamored with chatbots but to keep our focus on making te agents more efficient. Keep in mind that I live in the contact center world so I tend to spend a lot of time in that little slice of the overall customer experience.
When we look at NLP and machine learning, which are both members of the AI family, it’s important to recognize that, while there’s certainly excitement around automating everything, that also takes a ton of work, time, and data. What you’re more likely to continue to see are applications with these technologies built in that continue to learn and make significant gains in efficiency, slowly automating some of the low hanging fruit in the process.
This is definitely an exciting period where we’re training machines to get better at certain functions. Hopefully this helps you recognize where AI is actually making a difference as you evaluate tools and technologies in the CX marketplace.
Jeremy Watkin is a contact center veteran turned CX leader. He is an avid learner, and is constantly giving back to the CX and Customer Service community through his writings. You can see his work featured on Customer Think, Customer Service Life, and now CX Accelerator!