Knowhere’s Chatbots Have Beauty On Their Artificial Brains
Anybody can put up a website these days, but not anybody can make it deftly interact with customers. Chatbots offer the promise of advancing small businesses’ abilities to provide top-notch customer service. In 2016, Frederik Schröder co-founded Knowhere to develop AI-powered chatbots for consumer-facing companies for that purpose. The Hamburg-based company’s 12-person team strives to design chatbots that can become integral parts of digital customer communications. In a wide-ranging interview with Beauty Independent’s German edition, Schröder explains how Knowhere’s chatbots work, the time-saving advantages of machine learning and beauty of artificial intelligence.
AI and chatbots are buzzwords gaining traction in the industry, but they aren’t particularly new. When exactly did chatbots debut?
Chatbots have been around for a long time. The first draft of a communication design was created as early as 1999. At that time, this had nothing to do with AI, but was more a tool made for interaction on websites. Since then, however, much has happened and global companies such as Google have made a significant contribution to AI progress. Thanks to them, we have been able to implement chatbots for a few years now that are not only nice user-interface features, but will help improve customer communication in the long term.
A milestone year was 2016, when you could experience chatbots not only on a website, but also on Facebook Messenger. This made many people aware of the topic. There are currently around 300,000 Facebook Messenger chatbots, compared with 10,000 in June 2016.
That’s a rapid evolution. What role do chatbots play in companies today?
You can describe them as digital assistants. They are another team member in customer communication to automate repetitive questions and take over the task of answering the same questions again and again that can be sometimes very exhausting. Employees are thus relieved of their workload and can focus on more complex questions.
Customers have different intentions, and it’s not always about returning goods. What is the best way to subdivide the AI chatbot fields of application?
We divide chatbots into three customer journey stages. The first is the marketing: In this stage, the user can find out what kind of digital storefront it is and what they can buy there. The second stage is sales. Here, the user has a more targeted buying interest, which means they want to know, for example, the prices and how long the delivery times are. And the third stage is the customer service: I have bought a product, I am an existing customer, and there are questions about my account or my orders.
In what industries are chatbots the most popular right now?
They’re particularly prevalent in industrial e-commerce, insurance/financial services, and public institutions. Knowhere has already developed chatbots—automatic customer service chat systems—for more than 100 companies in the aforementioned fields (i.e., Geberit or Velux).
What makes them artificially intelligent?
First of all, the chatbot can also be stupid, of course—if you use it only as a guided dialog, let’s say, and customers can choose from multiple-choice answers. The answers of the chatbot are pre-determined and, therefore, the chatbot is not intelligent. Chatbots featuring artificial intelligence can understand free text queries. A chatbot can assign an inquiry to an intended question and learn by itself through skillful follow-up questions, thus opening up new subject areas and answering more and more queries automatically.
Sounds like a complex and time-consuming development process.
The greatest challenge is not the technology, but the communications. In order to be able to create a chatbot that can really understand the customers, our team of customer success managers makes sure that the chatbot works as successfully as possible from the start. They work with the company employees who are constantly communicating with the customers and, thus, know the queries best. Often, they are from the customer service, marketing, social media or sales departments. That’s the core work. The technical part such as the self-learning part is handled by our product. You don’t have to worry about that.
What is the first step for implementing a chatbot?
You have to find out what people ask most often. The FAQs on the website are sometimes a good starting point. Based on this, we create a mock-up. This is a video where you can see what a chat process can look like without the need to program anything. That would be the first consultation step. If you decide to work with us, it’s always for one year during which there’s an onboarding workshop. We visit the customer, and we discuss intended questions in detail. Within six weeks, the chatbot is set up and can go online. After going live, it is continuously optimized, learns new things and suggests new topics. During this year, we will be able to gradually increase the automation rate so that the chatbot will end up with a very high degree of automation.
Sounds smart, but what happens if the chatbot does not understand anything that it’s being asked?
In this case, the query is sent to a service employee. There is not only an automated chatbot that does everything on its own, but also a person for the fall-back option. Usually, they are called if the requests are too complex. On a website, a user is directed to query the e-mail address, and the request can be forwarded directly to the support staff. This standard option is much preferred because it works regardless of opening hours and availability. During the opening hours, the conversation can be transformed directly into a live chat.
When does a chatbot make sense for a company?
Our AI chatbots understand individual queries, so it is important to have a minimum volume of support tickets each month. If it’s only t10 queries, you don’t need a chatbot. If it’s 1,000 queries a month, it makes more sense because every second query can be automated.
Consultations in beauty are an elementary part of the purchase decision. Can a chatbot do this?
Personalized advice such as recommending the right color for a foundation is currently not possible. What is possible, though, is to create a chatbot that inspires people. [For example,] how can I imagine this product live? The chatbot supplies me visual impressions of how something might look in real life. This approach can make it easier to make a purchase decision. Of course, you have to define in advance which questions your customers are concerned with. However, our product cannot yet work as an intelligent consultant who knows which color best suits another.
How low maintenance is your chatbot?
We offer a full-service approach. Our AI experts make sure that the chatbot always continues to learn along the right lines. The customer success management team primarily monitors the chatbot’s suggestions and verifies that the right ones are being implemented. If you want the chatbot to be even better, it needs maintenance. If you want to keep it at the same level of intelligence, it doesn’t really need anything.
Well, you can’t send your chatbot to school. So, how exactly does it learn?
The chatbot analyzes the questions that are often asked. In advance, it has already been trained somewhat, but it now notices that many questions come in that it does not know yet. It suggests creating a new topic. The Customer Success Management Team then checks it out first to see if it’s working before it is implemented. This means that the chatbot learns from the customer feedback rather than from the employees.
What can a chatbot do better than a person?
It can scale the support because it is available 24/7 and in real time on weekends and evenings. Also, the costs are less because they are limited to the installation and license expenses. Furthermore, many people can communicate with the chatbot at the same time, thus reducing the team’s workload. However, chatbots will not be able to completely replace humans as we will always be better answering more complex questions.
So, the perfect scenario is an interplay of man and machine, right?
Originally published in the German edition of Beauty Independent on Sept. 9, 2019.