Knowledge graphs in companies: Basics, architecture & benefits

23. September 2025

Knowledge graphs – also known as knowledge graphs – are one of the top trends in applied artificial intelligence. According to the Gartner Trend Report, graph-based technologies will be used in around 80% of all data and analytics innovations in 2025 (in 2021, the figure was just 10%). The use of knowledge representation is also increasingly being discussed in technical communication. The technology has long been used to model knowledge and map complex relationships. In combination with generative AI, the knowledge graph becomes an even more powerful tool for efficiently organizing knowledge, recognizing relationships and enabling intelligent applications. But what does this buzzword actually mean and what are the benefits of introducing knowledge graphs in companies?

What is a knowledge graph?

Knowledge graphs are used in technical communication to link complex product information in a structured way and make it usable for both people and machines.

A knowledge graph is a structured representation of knowledge in the form of a semantic network. Information is linked together via nodes (entities) and edges (relationships). This semantic structure makes connections between information visible and machine-readable.

This allows information silos to be broken down and company-specific knowledge – for example on products, components or functions – to be consistently modeled and reused. It does not matter whether the underlying data is structured or unstructured.

How does a knowledge graph work?

In brief: A company’s knowledge is first described in a clear, standardized form, such as an ontology. Data from different data silos is merged, harmonized and enriched with metadata. This semantic enrichment creates a uniform context in which information is not only stored but also linked to one another.

The result is a machine-readable knowledge network that can be used flexibly. Knowledge graphs can be used in companies for many purposes, for example for intelligent assistants, more precise semantic search queries, automated analyses or for the development of new data-driven services. The more data is integrated, the more valuable and meaningful the knowledge graph becomes, because it makes connections visible that were previously hidden.

How to create knowledge graphs

Describing a knowledge domain with an ontology

The basis for a knowledge graph is an ontology. “An ontology is the explicit, formal representation of the knowledge of a specific domain,” explains Dr. Martin Ley, Professor of Information Management at Munich University of Applied Sciences. The domain is the subject area that the ontology covers. The ontology defines which classes of objects exist in the domain and how these objects (nodes) relate to each other (edges). The objects can be enriched with additional information, e.g. source, time or format, via further properties.

Examples of possible classes of objects in our context of technical communication are products and their components and parts, but also repair information or spare parts. Typical relationships are that a product consists of components or that repair information belongs to a component. This information content can be represented in so-called directed graphs. These are relationships (edges) that have a specific direction.

To make this representation machine-readable, it is expressed in a specific description language such as RDF (Resource Description Framework).

Specialized tools are used to model such ontologies in computer science, which support the creation, structuring and validation of knowledge models. The finished models are then stored in a graph database, where they can be flexibly queried and expanded. As a partner of various providers, we advise our customers on the selection and implementation of the right solution.

Specific content in the graph database

Once the basis for the knowledge model has been created as an ontology, the information is exported from the various data silos, transformed and mapped as so-called “proxy objects” in a graph database. These representatives, also known as instances, are the concrete data that represent the abstract classes of the ontology. This enables relations to exist between the real data as well.

This networked information is an excellent basis for value-adding business and service processes.

Challenges in knowledge modeling and knowledge graph construction

“The challenge in modeling an ontology is to identify the classes and relations that are relevant for future applications. As a rule, existing metadata concepts, information models or nomenclatures are used and processed automatically,” explains Dr. Martin Ley. “When building the knowledge graph, it is also challenging that product knowledge is usually stored in many different places in the company for historical reasons. Product-related information is stored in the PLM system, technical documentation in a CCMS and information for sales in a CRM system, etc.”

This requires one or more conversion tools that automatically transfer existing information into the graph database for specific use cases. “Simple mapping is often used, as the information is usually structured or semi-structured. If the information cannot be mapped adequately, additional functionalities such as text mining or entity extraction may help,” says Dr. Martin Ley.

This is precisely where our PANTOPIX SPHERE platform comes in: It automates many of these conversions, integrates information from different source systems and provides reliable, semantically linked data for a wide range of applications on the basis of a central knowledge graph.

This networked information is an excellent basis for value-adding business and service processes.

Possible applications of semantic knowledge structures

Knowledge graphs incorporate different types of information from different areas of the company and offer a 360-degree view of the company. A wide range of application scenarios can be derived from this:

Semantic search: They enable better search results because the search is based on context and not just keywords.
Intelligent assistants and chatbots: These systems use knowledge graph technology to answer questions precisely and contextually.
Data integration: Different data sources are linked and standardized so that correlations become visible.
Better support: Service cases are resolved more quickly and in a more targeted manner because all relevant information is linked.
Knowledge management: Companies organize and use their specialist knowledge more efficiently.
Recommendation systems: In the form of apps, they support sales staff or customers directly in the selection of products that require explanation.

Application examples: Graph-based information in practice

One conceivable application based on a knowledge graph is an intelligent service assistant, such as the Service Copilot from ZEISS. This provides service technicians at all service levels with all relevant information such as repair instructions, spare parts or work values in a product- and context-specific manner. In the form of a ticket system, 1st-level support could also be informed about suitable problem cases from other customers in addition to the complete service case history and thus find suitable solutions more quickly. An additional AI chatbot makes searching through hundreds of documents such as manuals and troubleshooting more efficient. Clear source specifications ensure the quality of the AI response.

Another example of the creation and application of a knowledge graph-based approach is described in our white paper “Step by step to the knowledge graph”. Using the use case of a pump manufacturer, you will learn how we have transferred product knowledge in tables to networked information in a knowledge graph.

Advantages of knowledge graphs in companies

In technical communication, knowledge graphs are a central technology that enables companies to use their valuable information for numerous future-proof applications – instead of just managing it in different data silos. AI applications also benefit from semantically enriched information: Large language models (LLMs) are known to have weaknesses in terms of the factual accuracy of their answers and with hallucinating. However, the demands on the quality and timeliness of technical information are extremely high. The combination of knowledge graphs and LLMs raises both technologies to a new level.

The special thing about using a semantic knowledge network is its flexibility and scalability. It is possible to start networking valuable knowledge today with a small amount of information and systems and gradually integrate additional subject areas and information sources. This ensures that the investment in a knowledge graph-based technology is an investment in the future.

The use of semantic knowledge structures brings many benefits:

Knowledge is better networked so that connections between different pieces of information become visible
Decisions can be made faster and more confidently thanks to a well-founded overview
Data from different sources and systems can be merged
Various departments such as service or support can answer queries more quickly and in a more targeted manner
Processes such as recommendation management can be automated
Networked information makes it easier to develop new solutions and innovations and is an outstanding basis for a wide range of portals
Knowledge graphs are at the heart of data-driven applications such as recommender systems, dashboards or customer 360 views
Compliance and quality assurance are supported by transparent data

Authors: Sandy Hedig, Marketing Manager at PANTOPIX, in collaboration with Dr. Martin Ley, Professor of Information Management at Munich University of Applied Sciences.

Status: August 2025