This case study delves into the challenges faced by Dreamix during our rapid company growth, specifically in managing internal processes and information access. To address these challenges, we at Dreamix identified an opportunity to leverage advancements in natural language processing, particularly large language models like ChatGPT. Our objective was to develop a streamlined and user-friendly solution for our employees.
In this study, we explore how we utilized the OpenAI API to create a custom chatbot that seamlessly integrated with the Dreamix intranet system. This chatbot served as a comprehensive resource, offering both static and dynamic information about the company, projects, teams, and individual experts.
The Challenge
As Dreamix experienced significant growth, managing internal processes and accessing internal information became increasingly challenging. Fortunately, the remarkable advancements in natural language processing, particularly large language models, opened up an opportunity for us to offer our employees a more intuitive and efficient way to access this information, ultimately saving them valuable time and effort.
However, utilizing ChatGPT directly was not an option, primarily due to the system’s lack of knowledge about our internal processes.
The Dreamix Solution
By harnessing the OpenAI API and its powerful large language models (which are also used in ChatGPT), we successfully developed a chatbot capable of responding to natural language inquiries regarding our internal processes. To overcome the limitations of the existing OpenAI technology, we got creative and seamlessly integrated the chatbot with our intranet solution. This implementation led to the following achievements:
- Exposing the full volume of static information about the company and the processes in it, despite OpenAI’s context size limits.
- To address the issue of OpenAI models occasionally providing false information, we implemented several effective solutions. Our approach involved providing direct links to official sources of information within our intranet system, implementing techniques to detect and restrict false information, and placing disclaimers where necessary. These measures were implemented to ensure employees are aware of potential inaccuracies in the chatbot’s responses and to maintain the reliability of the information provided.
- By seamlessly integrating our chatbot with our live intranet database, we enabled it to securely and reliably execute queries while maintaining flexibility. This integration allowed us to provide not only static information but also dynamic updates on current projects, teams, and individual experts. As a result, the chatbot’s capacity to serve a broader range of information was significantly enhanced.
More simply put, we combined the enormous existing knowledge of OpenAI’s pre-trained model with the Dreamix-specific information in our intranet system.
The Results
Our internal chatbot has become a popular tool among our staff, offering quick and natural access to relevant information. By using the chatbot, our employees not only save time and effort but also have the opportunity to explore the remarkable capabilities of large language models. This not only provides practical advantages but also fosters creativity and original thinking, as people discover unexpected ways to leverage the chatbot. As a result, employee satisfaction is improved, and they gain a heightened awareness of the possibilities that generative AI brings to the field. By empowering our team with efficient information access and encouraging innovative approaches, Dreamix achieved positive outcomes amidst our organizational growth.
