AI chatbots in intelligent 24/7 service in Hamburg

The Lübeck-based Drägerwerk is also using the knowhere product. The AI chatbot is being used on its global Facebook channel, where (long) response times had not kept pace with the rapid rise in customer enquires. The company, which specialises in medical and security technology, opted for a self-learning chatbot that understands the enquiries based on natural language speech recognition, pre-classifies them and responds in real time. The automated enquiries have boosted customer service. “A customer waits on average less than 60 seconds for a response to their request. And this is 24/7 and 365 days a year,” said Anna-Soraya Hein, Social Media Manager at Drägerwerk AG & Co KGaA. The process had previously taken up to six days.

Initially, knowhere had a completely different target market in mind when it launched in 2015 with a travel app that made suggestions for ideal holiday destinations using image recognition and AI. When the app proved less successful than anticipated, after a year the team began to rethink their approach. The first AI chatbot came about playfully, but found a taker immediately. The start-up soon secured customers like Mediamarkt and Jägermeister. “Along with the digital agency la-red, we developed the ‘Jäm Bot’, a rapping chatbot for Jägermeister that allows Facebook users to send individualized rap videos to their friends,” said Robert Weber, Marketing Manager at knowhere. “The campaign won two nails at the ADC 2017.

While marketing bots of this kind met with great interest from companies, knowhere aimed higher: “Marketing bots are in general aimed at the short term. As so-called ‘click bots’ they are simple to use and very good for one case, but that’s it,” Weber explained. But the knowhere team was more interested in long-term solutions and began focusing on developing an AI chatbot solution for more complex customer communication. Their approach was to train the AI on the basis of a Natural Language Processing (NLP) system to deal with a company-specific set of questions. NLP aims to comprehend and process natural language using rules and algorithms to ease communication between humans and computers as far as possible.

Weber used the example of a bot when insuring a car to illustrate how this works. “The bot is able to differentiate between damage caused by an accident, hitting an animal on the road or by the weather and then classify it,” he pointed out. It even masters different ways of spelling and synonyms. If the customer writes ‘collision’ instead of ‘accident’, this does not present a problem. “Our AI chatbot solution deduces the facts from the context and learns from it. And it recognizes clusters,” Weber added. When reports on a particular issue proliferate, such as after a heavy hailstorm like the one in Munich a few weeks ago, the bot transmits this information to a human service team. The company is then in a position to react rapidly by placing a button on its website reading ‘For hail damage click here.’ “That leads to satisfied customers,” Weber explained. And the bot does not even have to understand the word ‘hail damage’. “But it recognizes the frequency, reacts to it, asks for an explanation and then learns from it,” Weber added.

The customers of knowhere are happy with their AI chatbots. But what about the customers’ customers? Don’t people prefer to communicate with human contacts? “Of course, everyone would prefer an intelligent human as a point of contact, someone who is able to answer all the questions comprehensively,” Weber acknowledged. But what business can afford a 24/7 human service? “We perceive a high level of acceptance for our bots, particularly for first-level support,” Weber noted. Even self-learning AI chatbots reach the limits of their ability at some point – and then transfer matters to their human colleagues. “And for this reason, no one has to fear that bots will replace humans. On the contrary, they work alongside each other,” Weber added.
ys/rm/pb