How To Apply Deep Learning To Improve Customer Service
How much can we expect a machine to learn from a human?
Is it realistic to believe a machine is capable of adopting human communication, actions, and responses to the point of replicating human behavior?
It’s not crazy to think entire businesses may eventually be automated and able to adapt to market changes based on how well machines can learn from their human instructors.
Think about the last time you called your cell phone carrier or Internet service provider to talk about your service. You probably had to use an automated voice answering system before you eventually spoke to a human. Many automated answering systems aren’t very intuitive. In fact, you probably have more than a few memories of yelling into the phone or speaking like a robot before you either finally reached the person you wanted to or hung up in frustration.
A recent Forrester Research study shows almost 80 percent of U.S. adults choose “valuing a customer’s time” as the most important sign of good customer service. It’s hard to feel like your time is being valued if you’re becoming more frustrated with every attempt. For every misroute or repeated voice command the customer experiences, each instance can erode a caller’s trust in your company’s service.
What is Machine Learning, or Deep Learning?
So, what exactly is machine learning? Global management and consulting firm McKinsey & Company has an excellent definition of machine learning:
“Machine learning is based on algorithms that can learn from data without relying on rules-based programming.”
Instead of being programmed to do a certain task, computers can be fed a series of information. The systems will detect whatever data is entered and will match the new data with the data set already in its memory. For instance, rather than teaching a computer how to draw a dog, programmers can load millions of pictures of dogs into the computer’s database so it will recognize any dog or sketch that resembles a dog.
So, why is this important?
Machine learning builds mathematical and logical algorithms inside a computer’s processing system. This allows the computer to learn, adapt, and respond based on the information it receives.
You may wonder, “Is machine learning really that important?” It is if you can use machines to enhance your client’s care and provide admin support for more mundane tasks, like resetting passwords, collecting contact information, such as email addresses and phone numbers, and even provide sales support. Having a reliable machine-learning system in place frees up live customer service agents to handle more complex conversations with customers or revenue-generating activities, like closing a sale. With 75 percent of people preferring to solve a tech problem on their own, the value of having an automated system that learns what people truly need in the moment is priceless.
This is only made possible if the machine-learning experience is efficient, nimble, and most importantly, hyper-accurate. A Harvard Business Review study article states that businesses incorporating machine-learning processes and systems into their everyday operations are experiencing both top line and bottom line performance increases. Study author H. James Wilson wrote, “Powerful machine-learning algorithms that adapt through experience and evolve in intelligence with exposure to data are driving changes in businesses that would have been impossible to imagine just five years ago.”
Machine Learning Helps Businesses Handle High Volumes of Clients
Machine learning allows businesses to engage with a high volume of clients on a daily basis. Each interaction creates its own set of data specific to both the client and the customer service process itself. Machine-learning systems can curate the data collected from each interaction and adjust the algorithms for even better accuracy. Now, does continual training of a computer lead to ongoing costs? Yes, but think of how much time is already being invested in a customer service personnel base that experiences high turnover, low expectations, and a stigma of being one of the least fulfilling jobs in almost every industry.
Imagine if a customer service machine-learning process can also use predictive analytics to suggest a repeated issue that the caller may be calling about again. Imagine if a customer starts a phone call with Tech Support and receives a personalized response based on what the system already knows about their previous interactions. How do you think the customer will feel about the company? It’s a faster, more accurate, and friendlier process that no longer feels cold and uninformed. Integrating a Natural Language Processing (NLP) system may even make the customer service experience one of the easiest ways for a customer to feel appreciated and heard by your company.
As with many tech sectors, machine-learning startups are the ones with the freedom to explore some of the newer frontiers. Google spends billions of dollars acquiring fledgling startups, some of which turn into headline-snatching mistakes, like acquiring Nest’s smart home technology valued at $3.2 billion but riddled with bugs. Of course, many of Google’s acquisitions are massive successes. Is acquiring scores of wild-card startups expensive and unpredictable? Yes, but in the case of Google, money is never the issue and when they hit paydirt, the windfall may be historic. One of the biggest winning acquisitions made by Google is their $500 million purchase of DeepMind, a London-based artificial intelligence startup whose acquisition signaled the beginning of AI as a major tech sector.
DeepMind has already helped Google with energy efficiency. Google operates data centers around the world and put DeepMind’s data and machine learning capabilities to use improving power usage efficiency, or PUE as it’s also called, in all its data centers. DeepMind adjusted the computer usage patterns and reduced Google’s power usage by several percentage points. That may not sound like a lot, but think about this: Google used 4,402,836 MWh of electricity in 2014, roughly the same amount of energy used by 366,903 U.S. family homes in one year. Reducing power usage by one percent every year will translate to hundreds of millions of dollars saved in energy costs… and all because a machine learned how to use data to make changes. Using DeepMind to save energy costs will pay for the acquisition of DeepMind, but that’s just the tip of the iceberg.
In fact, Google had probably some of the most efficient structures in place for its data centers, and DeepMind still improved on their design. In spite of excellent over-all efficiency, Google had inefficient cooling systems in place before integrating DeepMind’s AI learning. Imagine how much energy, time, and most importantly, money could be recouped or saved by implementing machine learning for everyday processes. Retail giant Amazon is also experimenting with machine learning for fraud detection, website marketplace optimization, and even packaging and shipping processes. Simply restructuring a conveyer belt by a matter of feet could save Amazon tens of thousands of dollars every year in wasted energy.
Microsoft is rumored to be building a machine learning tool suite called Open Mind Studio. The intent is to integrate Microsoft’s machine-learning technology across its product spectrum, including Outlook, Skype, HoloLens, and search engine Bing. At the core of Microsoft’s machine-learning efforts is its Cortana Intelligence Suite combined with its Azure Machine Learning Studio that can be used by businesses and developers for creating, testing, and integrating predictive analytics with user data. The more tech leaders expand the reach of machine learning, the greater the potential for businesses to enjoy the possibilities.
Over a dozen European banks have replaced their outdated statistical-modeling systems with machine learning. The results are eye-opening in some instances:
- 10 percent increase in new product sales.
- 20 percent saving in capital expenditures.
- 20 percent increase in cash collections.
- And, 20 percent decline in churn.
Part of the machine learning systems initiated by the banks are microtargeted models that can accurately predict which patrons will default on bank loans or cancel service with a built-in intervention process. Machines are now helping save humans from their own spending habits by predicting failure and intervening before it becomes a reality.