Augment + Automate: The Latest Approach to Customer Service
The evolution of customer service has been far from linear. In ancient times, if you had a problem with a product or service, you simply went back to the vendor and complained.
From there, the vendor would either argue or fix it or replace the item, or improve the service.
Not much changed between then and the first half of the 20th century. Finally, in 1965, the automatic call distributor was invented, which allowed customer service centres to have a single customer service line that could be routed amongst however many agents.
This revolutionised the call centre, making large-scale customer service centres not only possible, but affordable. Since then, the dot-com boom, the invention of the cloud, and the rise of instant messaging, have all made their mark on the customer service industry, further transforming it.
But nothing has been quite so revolutionary and transformative for the call centre as AI-powered augmentation and automation. These powerful new ways of working are reducing workloads, stress, and overheads, and truly streamlining the powerful contact centre of the future.
What’s the difference?
These days, automation is the buzzword on everyone’s lips. With the power of AI, it’s now possible to automate everything from customer service responses, to text/data extraction, to business processes...heck, you can even set up automation to read, understand, and analyse large blocks of customer feedback.
But automation unchecked can go awry. Look at Microsoft’s Tay Twitterbot, which lasted for 15 unadulterated hours before the company had to take her down for racial slurs -- but not before she offended feminists, Jews, and Caitlyn Jenner. Enter augmentation: essentially empowering humans with the smarts of AI to help them do their jobs better, quicker, and more effectively.
Here at ultimate.ai, that’s how we roll. When we work with companies to build their chatbots, we start by augmenting their human agents. This is a win-win for everyone -- but we’ll save all the amazing benefits of augmentation for later in this article.
The importance of the human connection
The idea of automation is fantastic -- it allows us to take a step back from those repetitive, time-consuming tasks, so that we can concentrate on what really deserves our time. But not so fast, Husain Bolt. As we just said, unchecked automation is simply bad business. Plus, people prefer talking with other people. The human connection is oh-so-very important.
Just think of your own experiences -- how often have you pressed “0” on the customer service phone tree to skip ahead and speak to a human agent? Exactly. Machines are super helpful. But humans are better than machines at plenty of things -- emotional connection, empathy, and problem-solving, to name a few. And these are key to creating memorable customer experiences.
The science behind augmentation
Augmentation works in a very similar way to automation. It uses the same principles and the same NLP model. Markus Rautio, our CTO, explains:
“When you are using our tools to train an AI model for augmentation, you are also preparing for automation. We use a supervised learning algorithm where humans create links between a real message written by a visitor to something that we call an ‘intent’, which is the general idea behind what the visitor wants.”
Lost? Let us explain. Take lost passwords, for example. This is an example of a customer intent that can be worded in many different ways. To determine the intent, the technology works by first splitting a customer inquiry sentence into separate words. The words are then “translated” into machine-understandable code using a specially trained algorithm that has been linked to a specific language.
The AI then determines if the message fits into its pre-programmed intents, and gives a human-understandable confidence number based on how sure the system is that the question matches its predicted intent. The human agent, then, makes the final call, selecting the best answer -- which has a bonus effect of training the AI in real-time.
Benefits of augmentation
We promised to go into the benefits of augmentation -- and there are many. Let’s take a look.
- First, you’ll speed up your agent response times. Here at ultimate.ai, our suggestion tool reads the customer question, uses NLP to process and understand the text, and then offers human agents a selection of what it deems to be the three best possible responses. All the human agent then has to do is click on the best response of the three. That’s much easier than either a) typing the response out themselves or b) selecting the response from a long list of responses to copy-paste.
- Second, your agents will start responding to your customers with better quality answers. When building the suggestion tool, we sit down with our clients and go through five to 10 of their most common customer service cases. Then, they identify the gold standard of responses for each. This means that the AI is trained to suggest only gold standard responses.
- Thirdly, like Tony Stark’s Jarvis, the tool builds confidence. Instead of feeling on their own and perhaps at times unsure of the right response -- particularly when customer messages are pinging in from every which direction -- they have support. And not just support, but clever support.
- The fourth benefit of augmentation over automation is that augmentation actually helps to train the AI. So, when the agents are presented with the three best answers for a particular common customer inquiry, the AI remembers which response they selected. Over time, the AI learns which responses are consistently chosen by human agents, and which ones aren’t, and it adjusts its suggestions accordingly -- readying it for eventual automation.
The augmentation process
Another benefit of augmentation is that going live with it isn’t as scary as going live with automation. You can be comforted with the knowledge that there will always be a human you trust working with the AI to keep your customers happy and satisfied.
But that doesn’t mean that setting up the augmentation process is a simple procedure. Considering that your most common customer service cases make up anywhere from 30 to 70 percent of your inquiries, augmentation is going to play a big role in your company. So to make it the game-changer you’re after, it’s key to follow these steps when setting it up.
Understand your chatbot automation goals.
Have a good think about how you feel an intelligent chatbot will work its magic for your company. What are the issues your company is currently facing? Is it growing too quickly? Are your overheads extreme? Does your team struggle to keep up with your customer inquiries? Are your customers demanding 24/7 real-time support?
You want to whittle your answers down and drill down to the single most important reason you want a chatbot. Once you’ve established that, set some achievable and measurable KPIs that will keep your chatbot automation project on track.
Map out your common customer service cases.
Take a look at your customer service cases, and work out what are your most commonly asked questions. Chatbots are most effective at answering these FAQs, leaving your human agents free to attend to the more complex cases.
Not sure where to begin? Why not start by asking your team? Since they’re the ones constantly answering the same questions over and over, they’re best positioned to give you this information. Call a meeting to find out your five to 10 most common customer service cases, and make sure you make note of the most helpful answers too.
Validate with data
Armed with your most common customer service cases, it’s now time to go through your most recent chat transcripts. Every time you come to an instance of one of your most common questions, take note.
You want to discover how many times these cases appear, firstly, so that you can estimate your target automation level. Secondly, you want to start collecting these example questions, which you can use to train your chatbot.
Design your automated experiences
As you go through your common customer service cases, you’ll notice that they’re not always resolved the same way. At this point in your automation journey, it’s important to work out the gold standard for handling each of your most common customer service cases that you’ve selected for automation. Some cases may require some back-and-forth with the customer. For those situations, you’ll want to use this time to design the ideal conversational flow.
Train your chatbot
Remember how in step 3, you collected examples of your most common customer service cases? Now is the time to put those to use. Chatbots learn like people do -- with lots of different examples of ways the same questions and answers can be worded. The more such examples you can provide, the better equipped your chatbot will be to read, understand, and answer your common questions worded any which way. Aim for 30 to 50 examples of how each question could be asked and answered.
Now you’re ready to deploy your agent-facing suggestion tool. Prepare for the fact that you may be met with some resistance -- some people are intimidated or even scared of AI and worry that robots will one day steal their jobs.
So sit your team down first, use our Tony Stark example, and reassure them that the suggestion tool is only going to make their jobs easier and less stressful -- and make them more effective at what they do.
Now you’re ready to go live with augmentation!
Stay in the loop
Subscribe to our email newsletter. No spam, just occasional insight from our experts.