Why Semantic Control?

The Enterprise Problem

AI has an enterprise context problem

Semantic Control is an enterprise capability for defining, governing and applying meaning across information, AI systems and business decisions. It makes concepts, relationships, evidence, rules, authority and permitted actions explicit, reusable and traceable. With these guardrails in place, AI and agents then work within a dynamic context rather than relying on a statistical likelihood that the correct and complete information is reflected in results.

While AI systems and agents access your organisation’s documents, data, policies, systems and knowledge to ground responses, they do not understand the content.  AI does not know what your business terms mean, which information is authoritative, how concepts relate, which policies apply, or what actions are permitted in a particular situation. It only estimates the likelhood these might be realated,

A model might recognise the word customer. Your organisation might distinguish between a customer, policyholder, claimant, beneficiary, account holder and authorised representative. Those distinctions affect decisions, obligations and appropriate actions.  The problem becomes more significant as organisations move from AI assistants that generate answers towards AI agents that participate in business processes. 

Semantic Control complements strong data governance and information management. It significantly enhances AI governance, security and risk management by connecting automated interpretation, decisions and actions at scale with explicit, human-readable guardrails.

Traditional governance cannot solve the enterprise context problem.

Most organisations have established governance disciplines. 

Data governance addresses data ownership, quality, access, metadata, lineage and accountability. 

Information governance addresses the management, use, retention and authority of information. 

AI governance addresses the risks, responsibilities and acceptable use of artificial intelligence. 

All are necessary, but a gap remains between them. 

AI needs organisational context in a form that machines use consistently. Definitions, relationships, classifications, policies and business rules often remain distributed across glossaries, documents, systems, databases and the knowledge of experienced staff.

Semantic Control addresses this gap by making organisational meaning explicit, governed and usable by AI. 

The Enterprise Problem

This is a new problem with an established foundation

The underlying principle is not new. Information Engineering emerged in the 1980s from recognition that organisations should design information around the needs of the enterprise rather than individual applications. 

Information Engineering Planning started with the business: its objectives, activities, information requirements, entities and relationships. As enterprise technology evolved, elements of this discipline became distributed across enterprise architecture, data architecture, metadata management, master data management, records management and data governance. 

AI brings the original principle back into focus. If AI is expected to interpret enterprise information, support decisions and eventually undertake actions, the organisation needs an explicit understanding of the information environment in which AI operates. 

The principle remains unchanged: 

Start with the Business. Define its Information and Meaning. Then Apply Technology. 

Industry Thinking is Moving towards Semantics

Semantic layers created to ensure consistent meaning in structured data for analytics have proven inadequate for enterprise-wide production applications. Enterprise AI requires the concept to extend further, connecting structured data with documents, content and organisational knowledge through shared semantics, relationships, provenance and business context. The requirement for semantic control is increasingly reflected in mainstream industry research. 

Gartner stated in a May 2026 press release that AI agents depend on context, including semantic representations of organisational data, relationships and rules. Their recommendation is significant: organisations should establish a context layer as part of their data and analytics infrastructure.  Gartner warns that neglecting semantic foundations increases the risk of inaccurate agents, wasted expenditure and governance vulnerabilities.  

McKinsey reaches a similar conclusion in its work on agentic AI architecture. One of its core principles is to “share meaning, not just data.” It describes a semantic layer between data and AI applications that represents business meaning in machine-readable form, supported by ontologies, knowledge graphs, metadata and governance. 

Forrester also points toward this broader evolution. Its June 2026 work proposes a layered architecture where semantics provides consistency, ontologies formally define knowledge, and knowledge graphs connect enterprise entities and relationships, so AI agents have sufficient context for reasoning

 IBM similarly describes semantic layers as an architectural mechanism for translating complex data into meaningful business terms while standardising definitions, relationships and business logic. 

The terminology differs, but the architectural direction is increasingly consistent:
Enterprise AI requires governed business context along with Governed Data. 

Semantic Control Provides the Missing Layer

Semantic Control establishes a governed layer between enterprise information and the AI systems using it. 

It brings together six elements: 

Semantic Control does not replace data governance, information governance, enterprise architecture or AI governance. 

It connects them at the point where information becomes interpretation and interpretation becomes action. 

From AI interpreting information to AI operating safely within context

Without shared semantic control, each AI solution risks reconstructing business context independently. Different applications may use different definitions, retrieve different sources, apply different assumptions or interpret the same policy differently.

Semantic variance arising from proabalistic AI is hard to eliminate and becomes difficult to govern as AI adoption expands. Semantic Control changes the architecture so that the correct business meaning becomes the default reusable enterprise capability, rather than something built or embedded separately within every AI application.

AI models and agents use governed definitions, relationships, authoritative sources, rules and controls that give sufficiently reliable results for many non-critical workflows. The objective is to understand the cost of errors and the risks introduced by probabilistic AI in critical business areas. It is to determine where AI cannot be allowed to interpret based on statistical prediction and where the enterprise requires control over AI that is supporting decisions and taking actions at scale.

THE EXECUTIVE QUESTION

Who governs what your information means? 

Your organisation already governs systems, data, access and business processes.  What happens when AI increasingly interprets information and acts on your behalf?

Secret Link