About Markus and Retresco
Markus Bäuerle is Chief Operations Officer at Retresco, a Berlin-based agency specializing in automated communication solutions. The idea behind it is to automate routine communication in various industries and companies and to provide platforms for text generation. There are two aspects of text automation to consider: Text analysis and the analysis of data that is not yet available in text form.
These two dimensions are included in chatbots. On the one hand, he must understand what the person talking to him wants from him and, based on this analysis, he must be able to generate an answer. Retresco has developed solutions for internal communication between employees using chatbot technology.
How do you use chatbots for internal company communication and knowledge transfer?
Originally, this project was started for a customer who wanted a solution to enable internal knowledge transfer within the company. This major service provider in the finance sector had to comply with many regulatory requirements and process descriptions that it wanted to communicate to its employees. The previous solution was a classic search, with which documents could be searched through in a structured way using search terms, but it was difficult to answer the specific and very different questions from employees in this way. To solve this problem, a chatbot that can respond to employees’ specific questions was considered.
How is such a bot designed and implemented?
The basic idea behind the implementation was that the answers lie in the company’s documents and process descriptions. The bot must be able to extract the answers from the document and deliver them to the employees. These chatbots are less about dialog and more about answering questions. In addition to answering questions, he or she should also be able to do something for the employee, for example forward him or her to a required form.
There are various options for implementation. On the one hand, there is the domain option, in which the chatbot searches through and prepares the documents itself at the moment of the query. This option is particularly suitable for world knowledge questions or open topics. For customers such as large financial service providers, however, the response accuracy of this method is not high enough. For this reason, most of their questions and answers were set in prefabricated rooms. The most frequently asked questions were recorded in search statistics and lists. This collection of questions and possible answers was used to create a set of rules for the chatbot.
User expectations of the chatbot are very high in this context. The answers it provides must be correct and the system must be easy to check. To avoid misunderstandings, the chatbot’s answer must reflect the question asked or an alternative answer option must be shown. Scores can also be used to calculate the probability of which question matches the answer options. The chatbot’s response paths are structured differently. Some branch out, while others require only one entry to get an answer.
Learnings
1) Chatbots contain both aspects of text automation: On the one hand, it must understand what you want from it and, based on this analysis, it must be able to generate an answer.
2) This type of chatbot can respond to specific questions from employees. This facilitates the transfer of knowledge for companies in which knowledge is distributed heterogeneously.
3) The basic idea is that the chatbot’s answers lie in the company’s documents and process descriptions. The bot must therefore be able to extract the answers from the document and deliver them to the employees.
4) The response accuracy for such chatbots must be very high. For this reason, most of the questions and answers were set in prefabricated rooms. Finally, a set of rules for the chatbot was created based on a collection of questions and possible answers.
5) For this type of chatbot, it is important to ask yourself: What business model do I actually have internally and how can I change things there? Customers need to understand how they want to use a bot and which bot is best suited to which purpose. As a chatbot provider, it is therefore particularly important to seek dialog with the customer at an early stage and provide the necessary information.
It’s best to listen to the podcast episode with Sophie Hundertmark and Markus Bäuerle for yourself. Have fun![vc_empty_space height=”40px”]
And if you don’t want to miss any more podcast episodes, then sign up for the mailing list at the bottom of this page.
