Knowledge graph
A way of storing information as a network of entities and the relationships between them — people, companies, places, and how they connect. Knowledge graphs give AI structured, checkable facts to reason over, and are widely used in fraud detection and analysis.
A knowledge graph records the world as a web of connections. Instead of rows in a table, it stores entities — a company, a director, a bank account, a country — and the links between them: owns, works for, is based in, paid. Laid out this way, the information forms a network you can trace through, following relationships from one entity to the next.
This structure is powerful for questions that are really about connections. Spotting fraud, for example, often means noticing that several accounts quietly trace back to the same person, or that a chain of ownership leads somewhere unexpected. Those patterns are hard to see in a spreadsheet but stand out in a graph. Several UK companies, including Quantexa, have built their businesses on exactly this kind of analysis.
Knowledge graphs also complement modern language models. On their own, models can be vague or invent details; grounding them in a knowledge graph gives them a set of explicit, verifiable facts to draw on, which helps keep their answers accurate and traceable back to a source.