Goal
Increase productivity by optimizing internal knowledge management.
Solution
Development of an internal chatbot based on the RAG approach that combines the possibilities of generative AI with specific company information.
Result
Employees receive relevant information much faster and more precisely, saving them valuable time in their daily business.
Challenge
How can the enormous potential of generative artificial intelligence be harnessed without compromising on data protection or risking a loss of control? Like so many companies, our customer, a medium-sized mechanical engineering company, was faced with this tension between willingness to change and skepticism.
As a result, the company shied away from the strategic, comprehensive introduction of GenAI solutions such as ChatGPT - the risk seemed too great, the added value too uncertain. A survey of employees provided the impetus to tackle the issue after all. A key finding of this survey was that knowledge management was perceived as inefficient.
More specifically: Employees estimate that on average 10-20% of their working time could be saved if they could find relevant information, for example on work processes and responsibilities, more quickly and in a better format.
Goal
The mechanical engineering company wants to use the capabilities of large language models (LLMs) in its specific application case – the optimization of internal knowledge management using a chatbot – and thus increase productivity and create more freedom for employees to focus on value-added activities.
Solution
For the systematic introduction of GenAI and the development of a customized chatbot solution for knowledge management, the company relies on eoda's security-certified (ISO 27001) process and empowerment approach.
In which areas are there the greatest information needs? What are the key weaknesses in current knowledge management? eoda's experts, together with the responsible parties on the client side, identified the precise requirements and objectives. A crucial aspect of this process is the identification and prioritization of relevant data sources. In the specific case of the mechanical engineering company, for example, the focus was on work instructions and process descriptions from the internal wiki software.
This internal information is key for companies to sustainably benefit from large language models and reduce the risk of misinformation or so-called hallucinations. eoda's chosen RAG approach combines the skills acquired by a language model during its training with the company-specific knowledge of the mechanical engineering company. For this purpose, eoda has completed the data management and the data connection to the language model.
What is the RAG approach?
The RAG approach is a natural language processing method that combines information retrieval with text generation. It uses a pre-trained language model to retrieve relevant information from company-internal knowledge bases or documents and integrates it into the text generation. This allows the model to provide more detailed and accurate answers by accessing a broader range of information rather than relying solely on pre-trained knowledge. The RAG approach thus improves the quality and relevance of the generated content.
In addition to the question of relevant data sources, the question of the appropriate analysis model is always important in the data science context – and in the GenAI context, the question of the appropriate language model. Considering the framework, requirements, and data sources, eoda selected the appropriate language model for the mechanical engineering company. Given the rapid pace of development in this environment, the eoda approach also offers the possibility of flexibly exchanging the selected model during use to ensure consistently high performance.
To make getting started with and using the GenAI solution as easy and intuitive as possible for the mechanical engineering company's employees, eoda developed a customized user interface for the chatbot and integrated it into the customer's existing IT infrastructure.
eoda then handed over the system to the customer for independent further development and support. This handover was accompanied by training sessions – also at user level and with a view to successful use, but also general awareness of the opportunities and risks of GenAI solutions such as ChatGPT and Co.
Result
The solution created by eoda was already being used regularly by over 40% of employees within the first month of rollout. In a repeat team survey three months after the rollout, over 70% of employees confirmed that they received relevant and accurate information more quickly and were therefore more productive. By working with eoda, the mechanical engineering company's management has unlocked the potential of GenAI – without neglecting aspects such as trust, accuracy, and explainability.
Learn how we can support you in developing your RAG solution. Learn more.
Get started now:
We look forward to exchanging ideas with you.
Your expert on Data-Science-Projects:
Lutz Mastmayer
projects@eoda.de
Tel. +49 561 87948-370
