iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
McKinsey’s guidance describes a knowledge graph as one way to give AI a connected, business-aware view of enterprise data—not as a standalone fix for data readiness. The graph links information across systems and can connect unstructured material, such as documents and extracted tables, to structured entities such as customers, contracts, products, and transactions. Making that useful in practice also requires shared definitions, data quality, permissions, traceability, and workflow changes.
What a knowledge graph contributes to enterprise AI
A knowledge graph represents entities and the relationships among them. In McKinsey’s framing, it is commonly used with an ontology to implement a semantic layer between raw data and AI applications. The ontology defines the concepts and relationships that describe a business; the graph connects real-world data from different systems according to those definitions.
This layer gives people and AI systems a shared, machine-readable account of what business terms mean and how related information fits together. Without shared semantics, different agents may interpret the same data inconsistently, a risk that can grow as more agents and workflows are introduced. McKinsey explains this architecture in “Building the foundations for agentic AI at scale”.
How the graph connects structured and unstructured data
Enterprise information is more than rows in databases. It can include source documents and files alongside derived artifacts such as extracted text, tables, images, metadata, sensitivity tags, and quality scores. McKinsey’s data-readiness guidance says these representations should remain connected to their original sources and to one another, with ownership, sensitivity, and permitted use made explicit.
#1 Best Overall
An enterprise graph can anchor those artifacts to core business entities—for example, linking a contract document and its extracted clauses to the relevant customer, product, or transaction. That is different from putting files in a searchable repository: search can find content, but reliable AI use also depends on context, structure, versioning, and traceability. See McKinsey’s “AI data readiness: The key to scaling impact”.
What the graph does not replace
A graph is one component of an AI-ready data foundation, not a substitute for the practices that make data safe and dependable. McKinsey’s architecture guidance also emphasizes consistent ingestion, shared definitions, default security and access controls, monitoring, stable interfaces, and a controlled execution layer for applications and agents.
Rank #2
- Data quality: Teams still need to assess and improve the accuracy and completeness of source data and derived artifacts.
- Governance and access: Ownership, sensitivity, permitted use, and fine-grained permissions must be defined and enforced.
- Lineage and observability: Systems need to preserve where information came from and make quality and use observable.
- Operating model and workflow design: Teams must decide who maintains definitions and controls, and redesign the workflows in which AI is used.
McKinsey describes both a single agent that uses multiple tools and data sources sequentially and specialized agents that collaborate through shared knowledge graphs and fine-grained data access. In either pattern, the graph works alongside—not in place of—the controls and operating practices around it.
What McKinsey’s bank example establishes
McKinsey’s 2026 article on Central Europe’s AI opportunity describes a global universal bank that used an enterprise agent factory to reengineer its software development life cycle. The effort centered on business rules, agent-ready artifacts, and a shared knowledge graph. The published description does not identify the bank, graph technology, or implementation partner, and it does not quantify results attributable to the graph. It therefore supports the point that a shared graph featured in this workflow redesign, not claims about a particular vendor or measured graph-specific benefits. The example appears in “Capturing Central Europe’s AI opportunity”.
Rank #3
How to assess a knowledge-graph approach
McKinsey’s recommendations are architectural guidance, not a guarantee that adopting a graph will improve a particular AI system. Its older architecture guidance recommends starting from use cases, building a minimum viable data capability for each, and iterating. It also describes graph databases as a type of NoSQL technology that can represent relationships flexibly as information models change; that general context does not identify the technology used in the bank example.
For an implementation, assess whether the design can:
Rank #4
- Define business concepts and relationships consistently across teams and systems.
- Link unstructured artifacts to relevant structured entities while retaining source references and version context.
- Apply permissions and sensitivity rules to the data and relationships available to each user or agent.
- Expose lineage and data-quality signals so teams can detect and investigate problems.
- Provide stable interfaces that let the target workflow reuse governed data without bypassing controls.
For broader context, McKinsey’s earlier articles are “Breaking through data-architecture gridlock to scale AI” and “How to build a data architecture to drive innovation—today and tomorrow.”
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsQuick Recap
Best Value
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
