Intelligent bots as a smart solution?

A bouquet full of possibilities

“Always stay in touch! “— Today, this motto no longer applies only to the countless stars, starlets, influencers or local politicians who inform us almost continuously about news from their lives on various channels. It also applies to all service-oriented companies that value customer proximity and want to perfect their customer journey. The old motto “click, buy and forget” has long been a thing of the past. On the one hand, because companies always want to keep their customers up to date with new products, current offers and promotions. On the other hand, because existing and potential new customers also require this communication.

Over the past three decades, the range of communication options available for this purpose has expanded dramatically. Traditional options, such as personal contact, letters, phone calls and faxes, were supplemented with emails, chats or social media posts. Sometimes the channels are even mixed up in a wild, unpredictable mix.

But how are companies responding to the new requirements in the area of customer communication? In addition to the simple offer of connecting via a special channel, are there other technologies and methods to make contact and communication as profitable as possible for both the customer and the company? Fortunately, this question can be answered with a resounding “Yes! “answer. However, it should be noted that each of the numerous communication channels must be considered individually in order to be able to develop tailor-made solutions for each of them. In principle, communication channels can be divided into two classes: synchronous and asynchronous. Customers who make contact via synchronous channels usually expect an immediate response: Calls should be answered immediately. Chat requests — via web chat or messenger app via smartphone — require a quick, ideally immediate answer. And of course, customers also expect to be served quickly and satisfactorily in personal contact, for example when visiting a branch.

Asynchronous channels include letters, faxes, but also emails or social media posts. Here too, reactions are of course expected, but they do not have to be immediate. Instead, a certain delay appears not only to be tolerated but generally even expected. For traditional letter inquiries, this waiting time is probably the highest (in some cases weeks pass before the answer is received). For emails, the accepted duration is significantly lower (more like days) and on some social media channels, an even faster response is expected (even hours).

Bots as helpers in synchronous customer dialogue

The transition between synchronous and asynchronous channels is smooth, especially on the threshold between chat and social media. In the following, the focus is on synchronous channels. Here, companies are faced with the challenge of meeting high customer and service requirements. They must have sufficient competent personnel available to be able to respond to customer inquiries not only adequately but also as quickly as possible.

Numerous companies are meeting this challenge by offering their customers a artificially intelligent contact provide. So-called bots are able to both make phone calls and enter into customer dialogue via chat.

In principle, dialogue on both channels — speech and text — is based on an almost identical principle. Only in the preprocessing of the input and in the preparation of the output are there major differences: During Voicebot has to do with mumbling callers or dialects in the input, the chatbot be able to deal with spelling mistakes and symbolic inputs (emojis). When outputting, the voice bot must in turn ensure that this is done very quickly, as even the shortest breaks during a telephone call are considered unpleasant. The chatbot, on the other hand, should be able to symbolically signal and transmit messages if desired.

As a result, each of these two variants of human-machine communication has its own special features. This article focuses on the chatbot and the opportunities and risks associated with its integration into customer dialogue.

The art of conversation

First, there is the question of how communication via chat actually works. Traditionally, the conversation here is initiated by the customer, who sends a corresponding request to the company. The contact is linked to the expectation of receiving a quick answer. Conceptually, it doesn't matter at this point whether the answering contact person is a human or a vending machine. A freelance contact person will accept the chat request, greet the customer in the spirit of the company and ask a specific question about how it started. Based on the customer's response, a dialogue then ensues, in which his concerns are defined in more detail and all framework data for fulfilling this request must be clarified.

The dialogs always follow the same steps. While the chatbot reveals information and asks further questions, the customer provides answers and — if necessary — makes selections. In addition, the chatbot draws additional information from backend systems and can make decisions based on it that are important, sometimes even elementary, for dialogue management. For example, when a customer inquires about the whereabouts of a eagerly awaited order, the bot asks for the associated order number and uses this to check the delivery status. If there is also a shipment tracking number, this can also be transmitted so that the customer can independently find out about the further progress of their order.

How does the bot tick?

In order to find out which technology enables chatbots to conduct dialogue intelligently, a look behind the façade is essential. Here, too, there are basically two variants. Either the entire dialog is manually designed so that the chatbot ultimately only has to process a dialog script and follows strict if-then logic. Or the bot relies on special machine learning algorithms to provide appropriate answers based on the broadest possible information base. In this case, it is often used by Artificial intelligence (AI) Speaking, even though the buzzword certainly applies to both variants of automated dialogue management.

Of course, both types each have their advantages and disadvantages. In the case of the scripted variant, the bot operators are faced with the challenge of predicting all contingencies in the dialog processes and planning adequate responses. The wording and logical process should not be based on internal company practices, but should, if possible, be conceptualized from the perspective of a customer who lacks extensive knowledge of internal processes and relationships.

On the other hand, anyone who relies on self-learning bots is faced with a completely different challenge. In this case, a sufficiently large amount of data must be prepared and processed from which the chatbot can feed its knowledge. For example, existing chat dialogs led by employees could be collected and analyzed. In post-processing, it would then be necessary to manually note whether the individual dialog steps are to be classified as correct or incorrect. From the data generated in the process, the algorithm can then learn how it should behave in comparable situations in the future.

If such a chatbot with automatically learned behavior is then used in practice, there is a chance that it — similar to the famous Münchhausen lying baron — will pull itself out of the swamp by its own hair. This means that dialogs conducted by him are saved and individual dialog steps are evaluated by employees in order to expand the database and improve the reliability of the bot. However, once the chatbot has insisted on always handling certain situations incorrectly, the misconduct learned — unlike the scripted version — cannot be easily prevented. Instead, the database must be enriched with examples of the desired behavior pattern until the bot changes its learned concept of “right” and “wrong.”

Optimizing customer communication

But how can chatbots be used sensibly in customer communication? And how should their service be dosed? Regardless of the variant chosen, the question is how to deal with misunderstandings and where the bot's communication limits lie. So when is the point reached where the chatbot can no longer solve a problem independently, but has to hand it over to a human colleague?

While bots don't know the end of the day and can generally be deployed 24/7, human service employees are often not available around the clock. It is therefore important to decide whether the chatbot will only be used if the transfer to a human colleague is possible or whether the dialogue may be interrupted outside service hours. In the latter case, the bot could refer to service hours or initiate a personal appointment with an employee. The decisive factor for this decision is the scope of self-services the chatbot can cover. Provided that a reasonably complete portfolio of frequently used services is achieved here, round-the-clock operation seems useful.

In order to avoid misunderstandings, ambiguities and ambiguities in dialogue with the chatbot, it is possible to refrain from entering free text and to provide the customer with a limited selection of answers instead. In this way, linear dialog management can also limit the bot's response options — and thus the effort for the operator. The selection options can be displayed in various formats. In addition, they can be enriched with graphical information — such as product images — to visualize the various options.

In summary, it should be noted that automating customer dialogue is an extremely complex area of application in which it is necessary to analyse and weigh up in detail what type of inquiries a particular approach can be implemented and used profitably for. Bots are one of several ways that companies can efficiently perform one of the most difficult tasks in the business environment, communication with customers. After careful analysis and taking into account the special features described above, bots can become a decisive factor in optimizing customer communication. Your integration into Customer dialogue results in significant savings of two valuable resources: time and money. While previously stressed employees are relieved of routine tasks and are gaining new capacities, customers are pleased to receive assistance that is as quick as it is uncomplicated.