Artificial intelligence opens up new ways of utilising internal knowledge resources
Making good use of internal knowledge resources is an ongoing issue. People often talk of a “hidden treasure trove of data” that needs to be unearthed. However, different file formats, opaque filing systems and inadequate indexing hinder access to information just as much as silo thinking, inefficient processes, complex organisational charts or parallel data storage. The resulting problems are all too familiar: Teams unknowingly work on the same task in parallel, projects take longer than necessary and results fall short of expectations. The search for knowledge costs companies a lot of time and money, which they are usually unable to quantify due to the lack of transparency.
Thanks to artificial intelligence (AI), light is now being shed on this chronic data darkness. To be more precise: with the help of generative AI, company-specific EnterpriseGPT systems take on the challenging task of making internal data accessible, preparing it, making it available in a targeted manner, and thus raising the company’s internal data processing to a new level. adesso explains the three typical phases in setting up such a system:
1. Developing an internal AI chatbot:
The first step is to develop a company’s own internal AI chatbot. This avoids the data protection problems associated with the use of external AI chat systems. Company-critical information does not leave the boundaries of the company IT, not even for the training of LLMs. Instead of sending the data to the provider, language models are used that are available for internal use as software-as-a-service from various providers, including Google, Microsoft and Aleph Alpha.
2. Development into a “domain knowledge agent”:
Setting up your own internal AI chatbot first also has the advantage that the models can be trained using your own practical data, such as documentation, protocols, presentations or manuals, i.e. the internal knowledge base. Company-specific knowledge is provided in the form of domain knowledge agents for various knowledge areas and user groups. Appropriate authorisation management ensures that the relevant information is only accessible to intended users. Typical examples of domain knowledge agents are the clear summarisation and classification of internal research projects or providing answers to employee onboarding FAQs in plain language.
3. Developing “process agents”:
While domain knowledge agents answer specific questions and provide references, process agents take the next step towards automation. A process agent can, for example, independently handle the dispatching of enquiries to IT or create a quotation form based on a customer presentation and then make it available in CRM. However, this advanced automation of workflows places high demands on IT processes and requires a deeper integration of the system into the company’s IT landscape.
