Do you remember “Herr Kaiser”? The cult advertising figure of Hamburg Mannheimer Versicherung still embodies the ideal of a serious representative: expert, smart and always an ear for the fears and worries of customers. Listening, understanding and competent advice – in terms of customer communication Mr. Kaiser was a real role model. Today, virtual helpers such as chatbots or language assistants are to follow in these footsteps. A guest contribution by Patrick Zimmermann.
It affects not only the insurance industry, but almost every industry: In the course of digitization, customers are increasingly keen to get in touch with companies quickly and easily. The increasing interest in “talking systems” is a logical consequence of this development. Chatbots can be classified in the category “Conversational User Interface (Coversational UI)”. These include services such as WhatsApp or Facebook Messenger as well as language assistants such as Alexa or Google Home. Chatbots can also be implemented as plugins on websites.
Customer communication via chatbot: The advantages are obvious
Chatbots enable companies to respond to customer expectations and offer a digital dialogue. The virtual assistants enable private and direct communication in real time. Users thus automatically receive advice on urgent questions or are redirected to information on the website. This also reduces the daily flood of emails in a company.
Chatbots show their advantages especially in these areas:
- Customer service: A chatbot ensures that the company is available 24/7. If customers ask questions, the system reacts immediately and either provides a suitable answer or informs them that a customer advisor will deal with the matter shortly.
- Marketing: Chatbots are an ideal environment to hold competitions or to advertise certain products.
- Sales: Chatbots can be used to implement e-commerce strategies, for example to sell products and services. Using the modern interfaces of the bots, companies map their goods and start a purchase process.
In the insurance sector, for example, it makes sense to communicate offers for selected insurances via a bot or to submit a claim report digitally.
Technical basics: How a chatbot works
Meanwhile there are numerous enterprises, which offer Tools for the production of a Chatbots – frequently even free of charge. Typically, a modular system is used in which the bot functions are added step by step. Besides the design of a functional user interface, the biggest challenge in the development of a bot is the integration into existing systems. Often corresponding processes such as the interfaces (APIs) to payment systems or databases with prices are missing.
A distinction must be made between a rules-based bot and a system controlled by Artificial Intelligence (AI). Rule-based chatbots are relatively easy to program applications that guide the user through a predefined, arbitrarily expandable set of questions. The conversation then consists of already defined questions of the user and answers of the bot. If the user asks a question that is not included in the question tree, he only receives a standard replica.
AI-controlled chatbots, on the other hand, are characterized by their ability to learn. Finally, they should recognize the user’s intention and answer his questions according to the situation. The technical framework is complex because a developer has to use algorithms based on machine learning.
From Oneliner to Dialogue
So for a chatbot to have conversations as naturally as possible, he must first learn to speak. This is done using the above-mentioned question and answer rules and the machine processing of natural language, in English “Natural Language Processing” (NLP). The guiding principle behind NLP is that every form of language – whether spoken or written – must first be recognized. It does not only depend on the individual word, but on the connection with other words, entire sentences or facts. The term NLP therefore combines different techniques, all of which have the same goal: to scan a text for clues in order to recognize the author’s intentions and react accordingly. This includes the identification of certain keywords and example sentences as well as semantic and context-sensitive analyses.
Some of the most popular providers of natural speech processing tools:
- Amazon (“Lex”),
- Google (“Natural Language”),
- IBM (“Watson”) and
- Microsoft (“LUIS”).
These services provide language models that enable a chatbot to recognize the user’s intentions more and more reliably. Using the general patterns of conversation, the bot provides answers to individual questions. The more data is stored in the system, the more accurate the answers are, since the AI is constantly learning with human help. As a result, the bot becomes more and more “talkative”, i.e. it is more and more able to carry out a natural conservation with suitable answers. Thus an oneliner à la “Ich Tarzan, Du Jane” becomes a “real” conversation between bot and user step by step.
An example: The user asks what the weather will be like tomorrow in Leipzig. The bot tries to recognize the intention through NLP (“Intent” = weather). If this is successful, the system checks the entry for “Entities”, i.e. significant additional information (Leipzig = city in Germany, weather = activate weather API, date = tomorrow). From this the bot concludes that the user wants a weather forecast depending on the parameters mentioned. It transmits the information to a suitable interface – in this case the Weather API. It retrieves the desired data and sends it back to the bot, which feeds it into the chat. The user receives the answer to his question.
How a chatbot processes a customer request
Also chatbots have a personality
A conversational interface serves to improve the interaction between man and machine. In the business context, this refers primarily to the communication between the customer and the chatbot as company representative. Depending on the field of application, it is recommended that the chatbot “incorporates” a personality that comes into play in the interaction with the user. For example, a chatbot can learn to communicate in the youth language through appropriate conversation models. This makes a conversation more entertaining and the bot looks authentic. The customer feels taken seriously and understood – this generates positive emotions. In this context, addressing customers via “you” or “you” also plays an important role.
Which personality is suitable depends on the self-image of the company, its products/services and the target group. For a bank or insurance company, where respectability is a trademark, a too relaxed tone does not meet the customer’s expectations. Especially the business areas, in which a close customer exchange takes place, can profit from a chatbot with personality.
Implement a chatbot: Important Learnings
The experience from numerous customer projects teaches for the construction of bots:
- An established chat channel should be available
It is advantageous if a suitable infrastructure already exists, i.e. the customer can already chat with the company. Ideally, the chatbots are the application of a well-known provider. This makes it easier to integrate a new bot.
2. API interfaces for data processing should be available
API first! The questions where the chatbot finds the right data matching the user’s request are essential. Only if an interface for the data query exists, the construction of a chat bot makes sense.
3. A suitable team and suitable use cases facilitate the start
The team should include a UX designer for conversations and an experienced chatbot developer as early as the concept phase. The risk of misinterpreting the technical feasibility is high. At the start, the focus should be on simple use cases – for example in the areas of helpdesk and service. These can be queries of status messages or error reports. In addition, it makes sense to automate the most frequent customer questions at the beginning and to supplement special cases later. Regular feedback loops with the client help to avoid mistakes or to eliminate them at an early stage.
Conclusion: Chatbots belong to the future
The demands customers place on companies in terms of constant availability and high service are not diminishing. People expect a prompt and personal response to their concerns. For this reason, customer communication will be much more digital in the future. Chatbots and virtual language assistants provide solutions to meet expectations. Conversational interfaces form the technical backbone of these systems.
By 2020, 75 percent of all companies will have a chatbot in messenger applications. (Patrick Zimmermann)
It is important that the chatbot is not the only connection to the company, but that a service employee is still available for customer inquiries. Thus, the customer always has a human contact person at his disposal who can help if the chatbot cannot give an answer. Added to this is the potential of these applications as an e-commerce channel. Chatbots can therefore contribute directly to the added value of the company in the future. These developments are emerging:
- The tools and platforms for designing and creating chat bots are maturing.
- Communication between customers and companies via chat is becoming commonplace.
- In the future, chatbots will offer the possibility to buy products directly in the application.
About the author:
Patrick Zimmermann, CEO & Founder Knowhere GmbH
Patrick Zimmermann is founder and managing director of Moin AI, a specialist agency for messenger marketing and the development of chat bots. Patrick has been working in the fields of artificial intelligence and machine learning for more than ten years.
Picture source: © fotolia (Header), pixabay, Knowhere GmbH/Moin AI