Generative AI has a major impact on how companies manage and share their knowledge. Whether in support, administration or logistics, semantic search and AI-supported systems increase efficiency, close knowledge gaps and help to improve decision-making processes. Paul Engels, senior consultant data science & GenAI at Sopra Steria Germany, shares three practical examples of better knowledge management - from Specialsitter, a state authority and Deutsche Bahn, illustrating the potential of the technology in practice.
Corporate knowledge is one of the most valuable resources, but many organisations struggle to fully exploit its potential. Information often remains hidden in isolated data silos in individual departments, which makes knowledge sharing and collaboration much more difficult. When experienced employees leave the company, knowledge gaps arise because valuable knowledge has not been documented or passed on. A corporate culture that actively promotes the exchange and documentation of knowledge is often lacking. Added to this is the daily flood of information, where employees struggle to filter out the really relevant information.
Generative AI has the potential to overcome these hurdles. It makes knowledge available exactly where it is needed - be it in documentation, onboarding or daily collaboration. The results of the recent Sopra Steria study “Disruptive potential: How generative AI is defining new business models” show just how quickly the transformation is taking place. Six out of ten of the decision-makers surveyed believe that the use of the technology will change knowledge management in the next three years.
Semantic search in knowledge research
Today, companies predominantly use traditional search systems, such as those commonly found in operating systems like Windows or in internal wikis and intranets. These systems often reach their limits when faced with a wealth of information and only deliver relevant results to a limited extent. Generative AI can solve this problem by better understanding the context of search queries.
In contrast to conventional systems, a semantic search understands the meaning of a query and delivers context-related results, even if the exact terms do not match. Supplemented by chatbots with a natural language interface, knowledge searches become even more intuitive. Employees can ask questions in everyday language and receive precise answers including sources. Time-consuming manual research is no longer necessary and relevant knowledge is available at the touch of a button - quickly, clearly and in a user-friendly way.
So far, only 40% of the decision-makers surveyed expect GenAI to change collaboration at an interpersonal, technical and organisational level, the study results show. However, a look at selected application examples at a service provider and in public administration illustrates how this technology can increase efficiency.
Knowledge search at Specialsitter: How GenAI optimises a company wiki
Specialsitter, a fast-growing provider of assistance and care services for young people with disabilities, was facing a challenge. The company wiki, which serves as a central source of information for specialists, had become too extensive due to its growth. Employees were struggling to find relevant information using a simple keyword-based search. This inefficiency particularly affected the onboarding of new employees and led to time-consuming queries to colleagues.
To improve the knowledge search, Specialsitter implemented a GenAI-based solution with the help of Sopra Steria experts. At its heart is a web application that works like a chatbot. Employees can enter their questions in natural language. The AI searches through the data and provides fact-based answers by drawing exclusively on existing sources. This approach - known as Retrieval Augmented Generation (RAG) - minimises the risk of misinterpretation or so-called AI hallucinations. Questions that cannot be answered with the available data are transparently marked as such.
The answer itself is not a simple extract from a document, but is formulated on the basis of all available data. This approach ensures that employees receive precise and comprehensive information. They can now enter questions such as “What rules are there for team meetings?” and receive a comprehensive answer, including sources, in a matter of seconds. This efficient question-and-answer system not only increases productivity, but also ensures a consistently high quality of support at all locations.
Knowledge management in public administration
Like companies, public administration is also looking for ways to manage knowledge more efficiently. In one state authority, generative AI is used to search through large volumes of parliamentary data - such as small requests. The aim is to make the administration's work easier by providing relevant information faster and more precisely. This search system is also based on a Large Language Model (LLM), which relies exclusively on existing sources and is designed to prevent the risk of misinterpretation. An additional feature is that the system not only facilitates research, but also supports text tasks such as summaries and rewording. This versatile functionality saves administrative staff time, which they can use for other activities.
How Deutsche Bahn is using generative AI to optimise internal processes
Deutsche Bahn is also using generative AI to optimise its internal processes. One key use case is the management of extensive regulations, some of which can be up to 1,000 pages long. These documents contain safety-critical specifications for operation, maintenance and on-board service. With the customised solution BahnGPT, a platform was created that offers employees quick access to relevant information.
By implementing a specially adapted language model in the backend, queries can be made in natural language. The answers are based exclusively on approved and verified sources, which ensures compliance with data protection and security standards. Thanks to the semantic search and the AI's ability to process complex queries, the day-to-day work of employees in information security, technical support and operations is made considerably easier.
Before the introduction of BahnGPT, searching through the regulations was often time-consuming and error-prone. With the new solution, employees can now obtain precise answers in a matter of seconds without a lengthy search through documents. This not only saves time, but also increases operational reliability. Deutsche Bahn plans to extend the technology to other areas of application in the future in order to optimise even more processes and increase the quality of services.
Conclusion: modern knowledge management for better decisions
Generative AI is changing the way companies and public administrations organise and use knowledge. The examples of Specialsitter, the state authority and Deutsche Bahn show impressively that the technology not only increases efficiency, but can also improve the quality of decisions.
From semantic searches to customised applications such as BahnGPT, generative AI opens up new ways of providing knowledge in a targeted and contextual way. The use of such technologies minimises the search effort, reduces knowledge loss and increases productivity.
Practical experience makes it clear that anyone who adopts this innovative technology at an early stage can secure a decisive competitive advantage - whether in business or public administration.