Digitization in customer service
Really get your customers' intentions right with AI
How the digital transformation of your customer service succeeds
Who hasn't had this happen to them? In a hurry, you read an inquiry only superficially and jumped on a keyword. And promptly sent a completely wrong answer.
For your customers, this is annoying. Like for that customer who set up a webshop and found a small error in the VAT calculation. Not dramatic, but obvious and easy to correct - if you had access to the code and could program. So she contacted the website builder's customer service and asked for the error to be fixed.
In response, she got an explanation of how VAT is calculated. When she said that she knew that, she got an explanation of how to set up the VAT in the web store. She knew that too, otherwise she could hardly have found the error. It went back and forth for a while, but the service employee didn't understand that the customer was pointing out an error in the code to her. Presumably, it has not been corrected to this day.
An AI can do more than keyword recognition
If humans already have difficulty understanding a customer's intentions, can an AI learn to do that? Yes, it can. Because the AI has an advantage: It has no assumptions in its head and is not under time pressure. It doesn't think "Ah, the customer is writing something about VAT - she probably doesn't know how to calculate it." Instead, she evaluates what an email actually says. And she analyzes the context instead of reacting to keywords.
Only 37% of the companies use automation to analyze emails and other text messages from their customers. The vast majority among them use keyword recognition with word lists for this purpose. In fact, only 20% use the capabilities of Natural Language Processing (natural language processing) or other AI-supported processes (Source). A real AI with natural language processing can do much more than search for keywords. It learns from training texts which words and phrases customers have used for their inquiries in the past. For ReplyOne, 30-40 emails per topic are enough for this. An AI also takes into account and understands the context: What other words appear in the email? How are they linked to each other? Instead of searching for keywords, semantic speech recognition is used. The language patterns found are linked to a list of possible concerns that was previously created manually.
"I have noticed an error in the calculation of VAT". is recognized by the context as a complaint, the details and background of which need to be inquired. "How do you calculate sales tax?" on the other hand, is classified as a question that can simply be answered with text modules. ReplyOne immediately suggests a suitable solution to the employee, for example a link to a help article on your website.
When classifying, ReplyOne's AI even indicates how certain it is about the classification. This is a valuable hint for employees to read an email with low confidence particularly carefully and to check the classification by the AI. After all, when several topics are mixed in an email, when irony and innuendo are involved, classification is difficult for the AI - just as it is for us humans.
"Our expectations in terms of optimizing service processes in written customer service while maintaining a consistently high level of quality in responding to inquiries have been fully met with ReplyOne."
Especially in eCommerce and sales, the intention detection by ReplyOne is a valuable help. ReplyOne sorts the requests according to intention: customers with the intention to buy are placed in a sales queue. These messages have a high priority and are answered as quickly as possible. After all, companies that contact a prospect within an hour of a request have seven times the chance of a sales meeting than those that take two hours. If the response comes a day later, they have 60 times less chance and thus little chance of closing (Source: Oldroyd et al. (2011), Harvard Business Review). Fast and reliable classification and prioritization of incoming concerns are a clear competitive advantage.
How simple can good customer service be?
Artificial intelligence relieves the burden on employees
Questions about a product that has already been purchased and requests for a copy of the invoice, on the other hand, are pushed to different service queues. The questions are answered quickly. The dispatch of invoice copies can be largely or fully automated. This allows you to make optimal use of your service and sales team's resources. The million-dollar offer from the supposed Nigerian uncle and the repeated contact attempts from a pushy web designer looking for customers are moved to the spam folder. This saves pointless spent working time of your employees reading and sorting emails. To date, 63% of companies still sort and assign incoming requests to employees or queues manually (Source).
Complaints are forwarded to the best employees, who deal with them personally and individually. This is because emotions are involved in many complaints. Customers prefer personal contact and 60% cannot imagine interacting with a robot (Source).
Assign concerns correctly
Reduce working time
provides the right answer
Benefit from a well-trained artificial intelligence
As always with artificial intelligence, the quality of the training data is critical to the deployment for ReplyOne. Every hour you invest in selecting diverse yet typical and clearly understandable customer concerns will pay off later. Equally important is the selection of categories for classification. If important categories are missing or if categories overlap, it can be difficult for the AI to classify the concerns. We are happy to advise you on the selection of training data so that you can use ReplyOne successfully.
Recognize the intentions of prospects and customers with ReplyOne and sort the inquiries accordingly. This not only speeds up your team, but also increases the quality of the responses.
E-mails and other text messages land in the right queues and are answered by the best qualified employees. With ReplyOne you use the working time and competence of your employees for the really important work: the support of your customers and prospects.
How simple can good customer service be?