High cost is an issue in customer service. Is automation the answer? If it were only that simple. Automation of customer service is promising. But only intelligent deployment of automation will help you achieve both employee and customer happiness.
Excellent customer service is vital. And it is expensive. Some may have hoped that the shift from calling to text messaging would make life easier for contact centers. It turned out not to be true. Even more so: now companies have to offer more options than ever for communication with their customers: next to calls, there is email, live chats, and social media. Omnichannel or multichannel customer service makes things more complicated and more expensive.
But there is hope, and it is called automation. Enterprises cut back on human agents’ costs and invest heavily in automation of the customer experience. Investing highly in automation will not deliver results automatically. It might even bring productivity and NPS down. Here are three reasons to rethink your approach to automation.
1. Automating end-to-end conversations is hard
It is almost common sense to talk about the automation of conversations. We want to be able to automate the process fully. And before handing over the conversation to a chatbot, you have to be sure the machine can handle the conversation as a whole. Yet, this approach is unnecessary and even harmful.
The standard approach in automation is to treat a conversation as a series of questions and answers. In order to interact naturally with the customer, the machine has to be able to produce questions and answers in response to the customer (text) and move the conversation forward to the next segment (flow). This is what chatbots do, and it works fine with highly predictable end-to-end conversations. It looks like this:
This approach has three significant disadvantages.
The first one has almost become a cliche: customers don’t always like it. Especially when emotions come into play. In the end, a conversation is not just an exchange of information. It is far more.
The second disadvantage is that there are only so many predictable end-to-end conversations. The rest is still up to human agents, at least with the current state of technology.
And even if conversations start with a machine, once they drop off the chatbot conveyor belt, they do not get back on. This is the third disadvantage: the machine is a one-way street. After the chatbot fails, a human being is responsible for carrying that conversation, supporting the customer to a satisfying resolution.
All this causes machines to reduce a limited amount of traffic for most customer services. If we say machines successfully resolve 20% of your conversations, 80% remains.
2. Repetition is in the snippets
What if you don’t look at a conversation as a flow or a series of questions and answers. What if you look at a conversation as just typed data?
Let me explain.
Every conversation in customer service is unique. But across multiple conversations, you can often identify similarities. There are frequently occurring sentences and phrases in all written output. If we look at the whole of a customer service operation in an organization, hundreds or thousands of agents interact with customers, writing all day. And now, all these conversations together have even more frequently occurring content. And we can zoom out even further: thousands of companies, millions of agents, billions of conversations. All with similar structures and patterns. All with these similar predictable text pieces. Unique, yet comparable human interactions: across industries, across customer service. I mean, we all say 'Hi there' at some point, right?
Here is the point: if we look at the level of specific questions and answers, we see low repetition and low predictability. But if we shift the approach and look at the level of sentences, phrases, and text pieces, we see high repetition and high predictability.
And if it is repetitive and predictable, it can be automated.
Stick to the old approach, and you may automate some conversations fully. In all other conversations, your agents will keep on typing the same short phrases again and again.
3. Agents are great, but not for typing text
Your agents are too valuable to write the same text over and over. And then again: It’s not the agent’s job to write text but to engage in a meaningful conversation. Allow your agents to do just that.
And that raises the question: What can you do to help this human interaction? Can you assist the agent in focusing on customer satisfaction to increase NPS? Can you help the agent be more productive, reduce handling times, and improve happiness? Yes, you can.
Machines are much better than humans at many things. Machines make fewer mistakes and no typos and know your customers and your product better than you do. Machines can access and share unlimited data within milliseconds. But machines can not replace your employees. People are great where machines still fail. Enhance your people. Give them great tools. Make them type less, make them happy. They are great, just not at typing text. With autocomplete and reply suggestions, for instance, agents will have to type much less. You might consider hiring better paid and well-educated staff. People who may not be great at producing text but excel at keeping the conversation going using technology.
Your employees will be thankful, your customers too.
Many thanks to Johannes Wolters, our product owner, for laying the groundwork for this post.
Continue reading? Read how smart automation may quickly reduce handling times.