Storm Reply implements a knowledge management chatbot for Audi

Italian software house Storm Reply, a Reply Group company specializing in innovative cloud-based solutions and services, is supporting automaker Audi in making access to internal information faster, more efficient, and more reliable with a chatbot, based on Retrieval Augmented Generation (RAG), that leverages the power of Artificial Intelligence.

"The chatbot - says Filippo Rizzante, CTO of Reply - greatly improves the daily work of employees. It is able to integrate data from different sources to keep the knowledge base up-to-date. Security was built in from the start: the chatbot architecture was developed in a Virtual Private Cloud to ensure data security and user privacy".

The chatbot, developed directly by Storm Reply, makes it easier and simpler for all Audi employees to get answers to inquiries regarding a variety of areas: from project documents, to risk assessments, to contacts of people inside the company. The tool only provides answers that are relevant and relevant to the company: if there is not enough data available to process an appropriate answer, the chatbot tells the user instead of constructing an answer that is not completely correct.

This innovative chatbot, implemented through the Amazon SageMaker service, leverages Retrieval Augmented Generation to optimize the quality of responses. Indeed, this approach allows responses to be made more accurate through prompt engineering without the need to change the underlying generative model. With this tool, Audi's employees can have easy, fast and secure access to information, increase productivity and the quality of work and decisions, as well as make internal communication more efficient.

Michael Pawelke, Product Owner AWS Foundation Services at Audi AG, commented, "The chatbot developed by Storm Reply is an excellent example of how technology can be used to solve practical use cases in a very short time. Through Generative AI it breaks down barriers to knowledge gathering for our internal stakeholders and will require no further effort to be extended to other uses".