Recent advances in the field of artificial intelligence have the potential to fundamentally change the way companies analyze data. Facility Management is naturally also affected by this.
AI-powered Large Language Models (LLMs) can serve as business intelligence platforms, efficiently analyzing and presenting unstructured data such as internal company contracts, documents and reports in an engaging way. In the form of chatbots (think ChatGPT or Bard), they can enable more natural communication with users in areas like Facility Management and thus speed up many processes. In addition, their application to charted data allows for precise predictions and analysis, which are essential for optimal resource allocation and operational efficiency in Facility Management.
Despite the promising fields of application of AI-supported chatbots such as ChatGPT or Bard, it is important to ensure the security of sensitive company data. Numerous cases show how important it is not to upload internal company information to the cloud and to take appropriate security precautions.
This is why we have joined forces with our partner company Innomatik to kick-start the project “Chat4BI – Chatbot for Business Intelligence”. The project, which is funded by the Hessian State Chancellery in Germany, aims to identify possible solutions for the secure and reliable use of AI in the form of chatbots for sensitive data in companies across all sectors. We achieve this by developing smaller LLMs that can run on company-internal servers.
Over the course of 2024, we will develop joint prototypes in collaboration with the Innomatik team in order to evaluate existing database systems using natural language.
In doing so, we will primarily focus on
- connecting local LLMs to databases,
- evaluating these LLMs, and
- taking existing authorization concepts into account.
What would that look like?
An initial prototype could, for example, be a search by natural language input, as you know it from chatbots available on the market:
“How many free workstations are available on the 2nd floor?”
This would mean that the user would no longer have to use our evaluation engine in speedikon C, but could use natural language to retrieve the required information.
You can find more information on the general project website for funded projects in Hesse: https://www.lidia-hessen.de/
And, of course, we will also keep posting updates for you here and on our LinkedIn account.
