An enterprise knowledge graph is a map of how a company's information connects: which ticket is about which product, which bug caused it, which code change fixed it, which document explains it, and who worked on each. It is a part of some enterprise search products. A graph, in computing, is a set of things, called nodes, joined by links, called edges. Here the nodes are tickets, bugs, code changes, documents and people.
The graph is built automatically. The software reads the company's tickets, its bug list, its code changes and its documents. It then finds the links between them. Some links are written down: a ticket may mention a bug number. Other links must be inferred. For example, two tickets may contain the same error message, or a document may describe the same feature as a bug. To find text with the same meaning, the software often uses vector embeddings, lists of numbers that represent the meaning of a piece of text. The links are stored in a graph database, which is built to follow links quickly.
Here is a typical case at a software company in Chennai that sells billing software. A customer writes to support through the help desk: "Exporting an invoice to PDF fails when the invoice has more than 100 lines." This becomes ticket #5521. The knowledge graph finds six other tickets with the same error message from the past week. It links all seven tickets to bug #882 in the engineers' issue tracker. It links the bug to a code change in the PDF module, made nine days earlier. And it links that change to the engineer who made it, and to the help page about exporting invoices.
Now both teams see the same picture. The support executive sees that a fix is planned for Friday's release. She replies to all seven customers at once, with the date. The engineer sees that seven customers are affected, not one, and he raises the bug's priority.
Without the graph, the support team and the engineers keep separate records. Support does not know that an engineer is already fixing the problem. The engineer does not know how many customers it affects. The graph puts the customer's ticket and the engineer's bug in one connected record.
The knowledge graph is different from plain search. Search finds the documents that match the words of a question. The graph records how the documents are related. So an AI assistant can answer a question that crosses several systems, such as "Which customers are affected by this code change?"