Knowledge is crucial for organisational success ‒ consequently, knowledge sharing is of great importance. This is particularly challenging for multinational organisations: In addition to the known barriers to knowledge sharing at an individual, organisational and technological level, language barriers and cultural differences can also negatively impact knowledge sharing. While language barriers are often underestimated and cannot always be recognised as such, cultures with major differences in cultural characteristics tend to have a more challenging time with knowledge sharing. Therefore, successful knowledge sharing demands not only an adjustment of expectations on both sides, but also an awareness for the other culture and culture-specific behaviour. Even more, it seems that organisational culture can soften the impact of national cultural differences, particularly if a common culture can be established. Recent technological advances in the field of large language model based chatbots opened up new opportunities for intercultural knowledge sharing. They provide easy access to information by allowing users to interact with it not only in natural language but also in a variety of languages, regardless of the original language of the information, and can be linked to an organisational knowledge base to become a “domain expert”. And although culture cannot be conveyed through information systems, LLM-based chatbots can help understanding cultural differences better. Despite these promising features, the potential of using LLM-based chatbots for intercultural knowledge sharing has never been investigated in detail, much less tested with real users.

This master thesis explores how a developed chatbot based on a multilingual large language model that acts as an additional interface to an organisation's knowledge base can improve knowledge sharing between two national-culturally different entities within an organisation. A mixed-methods approach using literature review, interviews and focus groups allowed the identification of concrete knowledge sharing barriers within the organisation. Validated ideas were implemented in a prototype, which was evaluated with employees using realistic work tasks. The findings illustrate that a multilingual chatbot equipped with organisational data such as standards, internal knowledge bases and information about current and past projects can effectively improve knowledge sharing within two national-culturally different entities of an organisation.

The results suggest such a solution to be a viable approach for improving knowledge sharing, however, further research is needed to validate a broader applicability.