VISEON: Semantic Intelligence Glossary

Agentic AI

AI systems capable of autonomous decision-making and action-taking to achieve goals, using tools and resources independently while adapting to dynamic environments without constant human intervention.

Also known as: AI Agents, Autonomous AI, Agentic Systems

Agentic Catalogue

Agentic Catalogue: A unified, machine-actionable knowledge graph specifically engineered for autonomous AI discovery, interaction, and context retrieval. Evolving beyond traditional catalogues built for human browsing or basic data indices, an agentic catalogue consolidates multi-domain enterprise footprint variables (including ERP, CRM, MDM, and proprietary Intent Layers) into strict Schema.org-compliant structures. By leveraging Semantic Personification, it encodes not only structural records, but a brand’s signature voice, pedagogical methodologies, and mechanical constraints, allowing external AI systems to parse, trust, and precisely represent the entity.

Also known as: Digital Catalogue, Semantic Catalogue, Agent Ready Graph

Agentic Commerce

AI agents perform complex, multi-step tasks such as shopping: comparing products, negotiating, and making buying decisions, including payments, within automated workflows, often resulting in disintermediation of traditional e-commerce platforms.

Also known as: AI Search Channel, AI Shopping

Agentic Commerce Protocol

An open, machine-readable orchestration protocol that governs multi-step interactions between autonomous software agents and service environments across the “Transact” pillar. ACP transforms standard digital interfaces into agent-actionable environments by formalizing compliance rules, temporal pricing structures, and real-time fulfilment data (Available To Promise/ATP). Backed by modern enterprise infrastructure and semantic data platforms, ACP ensures that transactions are validated dynamically through the knowledge graph, shifting interactions from manual checkouts to fully autonomous, zero-neuron Discovery to Dispatch workflows while maintaining strict merchant oversight. See ACP as an integration protocol, ‘EDI for AI’. Supported by payment infrastructure providers like Visa (Intelligent Commerce), Mastercard (Agent Pay), and PayPal (Agent Toolkit).

Also known as: ACP, Agentic Payments Protocol, Discovery-to-Dispatch Protocol

Agentic Data Platform

A bimodal data architecture (ADP) designed to serve structured, high-fidelity knowledge to both humans and autonomous agents. It utilises Semantic MDM, GraphRAG, and MCP protocols to resolve digital obscurity and enable Agent-to-Agent discovery, negotiation, and commerce.

Also known as: ADP, VISEON Platform

Agentic RPA

Advanced robotic process automation where AI agents make autonomous decisions within automated workflows, enabling self-directed process execution.

Also known as: Agentic Robotic Process Automation, Intelligent RPA

Agentic Search

A governed information-retrieval and interaction architecture through which humans or autonomous agents can navigate, verify, and use structured knowledge using natural-language requests.

In VISEON, Agentic Search is controlled by a Stateful Behaviour Tree rather than handed directly to a language model. The SBT identifies the semantic task, consults validated state and governing contracts, and selects the appropriate execution path.

That path may be:

• an exact knowledge graph or KV answer;
• deterministic relationship traversal;
• an authorised tool or source-system operation;
• GraphRAG evidence retrieval;
• or bounded AI synthesis where language generation is genuinely required.

The language model does not own routing, state, evidence authority, or permissions. Responses can retain evidence documents, graph node identifiers, relationship identifiers, and execution provenance so that exact answers can be distinguished from retrieved or synthesised answers.

Also known as: AI Agent Search, Governed Agentic Search, Semantic Interaction

Agentic Web

An evolution of the World Wide Web where AI agents autonomously discover, interpret, and interact with web content and services on behalf of users or organizations. Built on protocols like MCP (Model Context Protocol) and NLWeb, enabling agent-to-agent and agent-to-service communication through structured data formats like Schema.org and JSON-LD.

Also known as: Open Agentic Web (Microsoft terminology)

Status: Emerging concept (2025), not yet standardised

Key difference from Semantic Web: Semantic Web makes data machine-readable; Agentic Web makes services agent-actionable

AI and Data Engineering

The application of artificial intelligence and data engineering principles to create innovative, scalable solutions that drive business growth and operational efficiency.

Also known as: AI Engineering, Data Engineering, AI/ML Engineering

AI Discoverability

The capability of content to be found, understood, and recommended by AI-powered search engines and generative AI systems. Brands without AI discoverability face digital obscurity in the age of generative search.

Also known as: Generative AI Visibility, AI Search Optimisation

Analytics

The systematic computational analysis of data to discover patterns, extract insights, and support decision-making across business and technical domains.

Also known as: Data Analytics, Business Analytics, Analytical Methods, Data Analysis, Quantitative Analysis, Statistical Analysis, Predictive Analytics

Augmented Search

Search systems enhanced with AI capabilities including natural language understanding, context awareness, and intelligent result synthesis through semantic search and generative AI.

Also known as: AI-Augmented Search, Enhanced Search

autoMagically

Automation that happens seamlessly and effortlessly, combining automated processes with intelligent orchestration. Essential for hyperautomation and agentic RPA implementations.

Also known as: Automagic, Automatic and Magical

Blackboard

A governed runtime state structure shared by the branches and leaves of a Stateful Behaviour Tree.

The Blackboard retains validated information needed to continue an interaction coherently, such as the active task, selected method, committed user intent, current knowledge entities, retrieved evidence, provenance identifiers, unresolved questions, and the interaction surface currently being presented.

Unlike a conventional chat history, the Blackboard does not treat assistant prose as an authoritative source of state. Information is admitted, validated, updated, and consumed according to explicit runtime contracts.

The Blackboard provides interaction continuity and semantic state. It does not replace authentication, authorisation, or role-based access control. Underlying source systems remain authoritative for user identity, permissions, and data access.

Also known as: SBT Blackboard, Semantic State Store, Governed Interaction State

Business Intelligence

Technologies, applications, and practices for collecting, integrating, analysing, and presenting business information to support decision-making.

Also known as: BI, Analytics, Business Reporting

Category Theory

A mathematical framework for describing abstract structures and relationships, applied to knowledge graph architectures for compositional coherence. VISEON applies category theory principles to ensure knowledge graphs maintain mathematical consistency and composability.

Context Gravity

The tendency for governed, authoritative, and well-connected semantic context to attract retrieval, interpretation, and agent attention toward the correct entities, relationships, definitions, and evidence. Strong Context Gravity reduces ambiguity by making the intended meaning easier for humans and AI systems to discover, select, and retain.

Context Gravity is the semantic-era evolution of Data Gravity: where Data Gravity attracts applications and services toward accumulated data, Context Gravity attracts retrieval and agent reasoning toward authoritative meaning.

Also known as: Semantic Context Gravity, Contextual Authority, Contextual Attraction

Data Governance

Framework of policies, processes, and standards that ensure data quality, security, and compliance across an organisation. In the context of agentic commerce, data governance ensures knowledge graphs maintain accuracy and consistency, enabling AI agents to trust and act upon enterprise data with confidence.

Also known as: Data Management Governance, Information Governance, Data Stewardship

Data Influencer

Recognised thought leaders and content creators who shape discourse, trends, and best practices in data science, analytics, business intelligence, and AI domains.

Also known as: Data Thought Leader, Analytics Influencer, Data Science Influencer, BI Influencer, Data Community Leader, Data Industry Expert

Data Science

Interdisciplinary field that uses scientific methods, processes, algorithms, and systems to extract knowledge and insights from structured and unstructured data.

Also known as: Data Analytics Science, Applied Data Science, Computational Data Science, Statistical Data Science, Machine Learning Science, Data Mining

Deterministic Answer Path

An execution route that produces an exact answer without invoking a generative language model when the required facts, relationships, calculations, or rules are already available in governed systems.

A deterministic answer may be produced from a knowledge graph node, a relationship traversal, a validated KV contract, a calculation engine, a policy rule, or an authoritative source-system response.

Within VISEON DISCUSS, the Stateful Behaviour Tree checks for an exact deterministic path before admitting GraphRAG retrieval or bounded AI synthesis. This reduces latency, model cost, linguistic drift, and unnecessary inference.

“Zero-neuron” refers specifically to avoiding LLM inference. It does not imply that no computing resources are used.

Also known as: Exact Answer Path, Zero-Neuron Path, Deterministic Leaf

Digital Obscurity

The state of being invisible or undiscoverable to AI-powered search engines and generative AI systems due to lack of structured data. Digital obscurity is the new invisibility – brands without semantic markup are effectively invisible to AI-powered discovery.

Digital Twin

A digital twin is a virtual representation of a physical object, system, or process, continuously updated with real-time data to mirror its state, behaviour, and interactions. It supports knowledge graphs by providing a dynamic, data-rich model that integrates structured and unstructured data, enabling advanced analytics, simulation, and decision-making through interconnected nodes and relationships in the graph.

Generative AI

AI systems capable of generating new content including text, images, and responses based on training data and user prompts. Generative AI systems like ChatGPT, Claude, and Gemini are fundamentally changing how people discover information.

Also known as: Generative Artificial Intelligence, Gen AI

GIST – Greedy Independent Set Thresholding

A novel data selection algorithm, adopted by Google for search, developed by Matthew Fahrbach, that helps solve this issue by balancing data “diversity” (ensuring the selected data is not redundant) and data “utility“ (data that is relevant and useful for the task). GIST not only outperforms state-of-the-art benchmarks tasks, such as image classification, but it does so with a mathematical guarantee about its solution quality.

Also known as: GIST Algorithm, GIST Protocol

GraphRAG

A retrieval and evidence-assembly technique that combines graph relationships with relevant document or content retrieval to supply grounded context for an answer.

GraphRAG can identify connected entities, traverse meaningful relationships, and retrieve supporting evidence that ordinary vector similarity may overlook.

Within VISEON, GraphRAG is an evidence capability, not the runtime controller and not the source of every answer. The Stateful Behaviour Tree determines whether GraphRAG is needed, which graph and content sources may be queried, and how the retrieved evidence may be used.

Exact knowledge graph, KV, relationship, calculation, or policy answers are resolved deterministically where possible. GraphRAG is used where broader supporting context is required. Bounded AI may then explain or synthesise the selected evidence without becoming the authority for the underlying facts.

Also known as: Graph-based RAG, Graph-Guided Retrieval, Relationship-Aware Retrieval

Hyperautomation

Business-driven approach to rapidly identify, vet, and automate business and IT processes through orchestrated use of multiple technologies including AI and RPA.

Industry Trends

Observable patterns and directional movements in technology, business practices, and market dynamics that shape industry evolution and strategic planning.

Also known as: Market Trends, Industry Patterns, Sector Trends, Business Trends, Technology Trends, Market Movements, Industry Developments

JSON-LD

A method of encoding linked data using JSON, widely used for implementing Schema.org structured data. JSON-LD is the preferred format for semantic markup because it separates structured data from HTML content.

Also known as: JSON for Linking Data

Knowledge Graph

A structured representation of entities, attributes, and relationships that enables machine understanding and reasoning. Knowledge graphs form the foundation of how AI systems understand and connect information across the web. Schema catalogs can be derived from knowledge graphs.

Large Language Model (LLM)

AI model trained on vast text datasets to understand and generate human language. LLMs power generative search engines and increasingly rely on structured data and knowledge graphs for accurate information retrieval.

Also known as: LLM, Language Model

Language Model Optimisation (LMO)

The strategic discipline of governing and structuring an entity’s data — specifically its Nouns (Entities) and Verbs (Relationships) — to ensure maximum deterministic fidelity and citation accuracy within AI models and autonomous agents.

Technical Context: Unlike traditional SEO, which optimises for probabilistic search engine ranking, LMO focuses on the “Inference Layer” of AI. It utilises structured data protocols (such as JSON-LD, OSI protocol, and Knowledge Graphs) to move a brand’s identity from a “Variable” (calculated guess) to a “Constant” (verified fact).

By providing a Single Source of Truth (SSoT) through machine-readable catalogs, LMO reduces model hallucination and secures authoritative citations in agentic workflows. This infrastructure allows a brand’s identity to transition from a probabilistic variable to a verified constant, enabling qualified transactions and auditability for autonomous agents.

Also known as: LMO, Language Model Optimization

Master Data Management (MDM)

The discipline of defining and managing an organisation’s critical shared data — customers, products, suppliers, and employees — as golden records: a single, authoritative source of truth across all systems and processes. The semantic strategy equivalent for outward-facing AI discoverability applies the same rigour to machine-readable knowledge graphs.

Also known as: MDM, Enterprise Master Data

Mechanical Personalisation

The practice of translating an organisation’s distinctive operational logic, methodologies, interaction rules, and constraints into governed, executable runtime contracts.

Within VISEON, Mechanical Personalisation is implemented across the semantic architecture:

• the knowledge graph describes entities, concepts, methods, and relationships;
• semantic contracts define meaning, authority, constraints, and coverage;
• governed stores such as KV retain editable policies and approved language;
• and the Stateful Behaviour Tree executes the applicable rules through explicit branches and terminal leaves.

This separation allows an organisation’s behaviour to remain consistent without expecting a language model to infer operational rules from prompts, documents, or brand tone.

Mechanical Personalisation governs how the system behaves. Authentication and access permissions remain the responsibility of the connected source systems.

Also known as: Deterministic Brand Logic, Behavioural Semantic Control, Governed Runtime Personalisation

Model Context Protocol (MCP)

An open protocol that enables AI assistants to securely access data and tools through standardised server connections. MCP servers and tools allows AI systems to dynamically retrieve information from knowledge graphs and external data sources.

Also known as: MCP

NLWeb

Microsoft Research protocol for web-scale natural language understanding optimised for LLM ingestion through structured JSON-LD. NLWeb demonstrates how major AI research teams are prioritising structured data for training language models.

Also known as: Natural Language Web

Ontology

Formal specification of concepts, relationships, and constraints within a domain, enabling shared understanding between systems. Ontologies provide the semantic framework that allows different AI systems to interpret knowledge graphs consistently.

Also known as: Web Ontology

Open Semantic Interchange

An open framework for the structured exchange of semantic knowledge between AI systems, platforms, and agents. Open Semantic Interchange defines how entities, their relationships, and contextual metadata are serialised, transmitted, and consumed across heterogeneous systems using JSON-LD and Schema.org vocabularies — enabling interoperability between knowledge graphs, LLMs, and agentic infrastructure.

Also known as: OSI, Semantic Interchange

Pedagogical Branding

The systematic practice of encoding an organisation\’s signature teaching philosophies, conceptual frameworks, and distinctive transfer of knowledge into a machine-readable format. Rather than simply classifying factual brand data, pedagogical branding structures the rules, stances, and step-by-step methodologies unique to an author or institution, allowing external artificial intelligence systems and conversational agents to replicate and instruct via those specific frameworks without losing the creator’s intent.

Also known as: Educational Brand Alignment, Methodological Branding, Authored Concept Architecture

Python Package Index (PyPI)

The official repository for Python software packages, enabling distribution and installation of Python libraries. VISEON’s tools and integrations are available through PyPI for easy implementation.

Also known as: PyPI, PyPI.org

Provenance Contract

A machine-readable record describing how a response, decision, or action was produced.

A Provenance Contract can identify the Stateful Behaviour Tree route and terminal leaf used, the evidence documents retrieved, the knowledge graph nodes and relationships traversed, the deterministic engine or external tool invoked, and whether a language model contributed to the final expression.

This allows a consuming system to distinguish between:

• an exact deterministic answer;
• a graph-derived relationship answer;
• a retrieved evidence answer;
• a bounded AI synthesis;
• and an unsupported or unresolved request.

Provenance is retained alongside the answer rather than represented only as explanatory prose. This makes responses more inspectable, testable, and suitable for machine consumption.

Also known as: Answer Provenance, Evidence Contract, Execution Provenance

Retrieval Augmented Generation (RAG)

AI technique that combines information retrieval with text generation to provide more accurate and contextual responses. RAG enables generative AI systems to ground their outputs in retrieved information from external data sources such as document collections, vector databases, or enterprise knowledge bases, reducing the likelihood of hallucinations. Advanced protocols like GIST (Greedy Independent Set Thresholding) further optimise this by ensuring retrieved data is both high-utility and non-redundant, maximising the unique information gain within the model’s context window.

Also known as: RAG, RAG AI, VectorRAG

Resource Description Framework (RDF)

W3C standard for describing resources on the web through subject-predicate-object triples. RDF forms the foundation of the semantic web and knowledge graph technologies.

Also known as: RDF

Schema.org

Collaborative vocabulary for structured data markup on web pages, enabling search engines and AI systems to understand content semantics. Schema.org provides the standard vocabulary used across the web for describing entities and their relationships.

Search Engine Optimisation (SEO)

The practice of optimising websites to improve visibility in search engine results through technical, content, and structural improvements.

Also known as: SEO, Search Optimisation

Semantic Contract

A governed, versioned, machine-readable asset that defines the business meaning, authority, constraints, coverage, and permitted use of data or knowledge exposed through an interface, API, knowledge graph, or Model Context Protocol connector.

A Semantic Contract tells a consuming system not merely what fields or entities exist, but what they mean, which source is authoritative, how terms and metrics are defined, where coverage is incomplete, which relationships are valid, and which questions cannot be answered reliably.

Within a Stateful Behaviour Tree, Semantic Contracts are consulted before an execution path is selected. They help the SBT determine whether a question can be answered deterministically, requires evidence retrieval, needs clarification, must be delegated to an underlying system, or lies outside the governed knowledge boundary.

Semantic Contracts govern meaning and use. They do not grant user permissions or override the authentication and role-based access controls of underlying systems.

Also known as: Semantic Data Contract, MCP Semantic Layer, Governed Semantic Definition

Semantic Control Plane

A governed architectural layer that controls how organisational meaning, context, authority, provenance, rules, and machine-readable contracts are exposed to and used by AI systems and agents.

A Semantic Control Plane coordinates semantic contracts, knowledge graphs, Stateful Behaviour Trees, governed retrieval, deterministic engines, and bounded AI. It determines which source is authoritative, which execution path is permitted, what evidence supports an answer, and where the reliable knowledge boundary ends.

It governs semantics and machine behaviour, not user permissions. Authentication, authorisation, role-based access control, and data-access rights remain authoritative in the connected source systems.

Also known as: Enterprise Semantic Control Plane, Governed Semantic Control Plane, Semantic Decisioning Layer

Semantic Personification

The engineering practice of constructing a governed digital representation of an organisation, product, expert, or authored methodology that preserves not only factual identity but also distinctive language, reasoning methods, operational boundaries, and interaction behaviour.

Semantic Personification is implemented across several coordinated layers rather than being embedded entirely inside a knowledge graph.

The knowledge graph defines entities, concepts, relationships, and authored methods. Semantic contracts define meaning, coverage, and authority. Governed content stores retain approved voice and editable language. Mechanical Personalisation expresses operational constraints. The Stateful Behaviour Tree determines how those assets are applied during an interaction.

This architecture allows a conversational interface or AI agent to reflect an organisation’s recognised character and methodology without allowing a generic language model to invent its identity, infer its rules, or silently exceed its knowledge boundary.

Also known as: Governed Digital Personification, Mechanical Personalisation, Pedagogical Branding

Semantic SEO

SEO strategy focusing on meaning and context through structured data to help search engines understand content relationships and intent.

Also known as: Semantic Search Optimisation

Semantic Engine Optimisation (SEO)

The practice of architecting information, entity relationships, and topical authority so that reasoning engines can accurately understand a brand, its intent, and its commercial reality.

The Mission: To ensure the engine doesn’t just “find” your content, but fully understands it, eliminating the “understanding gap” that prevents accurate AI discovery.

The Semantic Core: Unlike legacy interpretations that focused on the mechanical retrieval of strings, Semantic Engine Optimisation is the process of building machine-readable intelligence, thereby becoming the evolved state of legacy Search Engine Optimisation

Also known as: Next-Gen SEO, Machine-Readable Architecture

Semantic Search

Search technique that understands user intent and contextual meaning rather than just matching keywords. Semantic search powers modern AI-driven discovery by understanding the relationships between concepts.

Also known as: Contextual Search, Intent-Based Search, Meaning-Based Search, Natural Language Search

Semantic Strategy

A strategic approach to organising and exposing business data using semantic web technologies (Schema.org, JSON-LD, knowledge graphs) to ensure AI discoverability and enable agentic commerce. Semantic strategy transforms fragmented digital content into a unified, machine-actionable catalogue that AI agents can discover, understand, and transact with, eliminating digital obscurity.

Also known as: Semantic Web Strategy, Knowledge Graph Strategy, Structured Data Strategy

Semantic Web

An extension of the World Wide Web that enables data to be shared and reused across applications, enterprises, and communities through standardised formats. Built on RDF and ontology frameworks.

Also known as: Web 3.0, Machine-Readable Web

SPARQL

Query language for databases that use RDF format, enabling semantic queries across knowledge graphs. SPARQL allows complex queries across distributed knowledge graphs using semantic relationships.

Also known as: SPARQL Protocol and RDF Query Language

Stateful Behaviour Tree

A governed execution architecture that routes an interaction through explicit semantic branches, conditions, capabilities, and terminal leaves.

A Stateful Behaviour Tree does not ask a language model to decide freely what should happen next. It evaluates the current request against validated intent, committed state, semantic contracts, available evidence, source authority, interaction surfaces, and operational constraints before selecting the capability permitted to answer.

Within VISEON DISCUSS, the SBT follows a governed hierarchy:

1. Deterministic — use an exact knowledge graph, KV, relationship, calculation, rule, or source-system answer where one exists.
2. Semantic — retrieve and assemble governed evidence through graph traversal, GraphRAG, content indexes, and semantic contracts.
3. AI — invoke bounded generative AI only where language synthesis, explanation, or natural conversation is required.

The SBT works with a Blackboard that retains validated state across turns. This allows the system to continue the active task, preserve prior decisions, obey the current interaction surface, and distinguish authoritative runtime state from ungoverned conversational prose.

The knowledge graph supplies meaning and relationships. Semantic contracts define authority and boundaries. Deterministic engines perform exact operations. GraphRAG retrieves evidence. The language model provides bounded expression. The SBT governs which of those capabilities is allowed to act.

Also known as: SBT, Governed Stateful Behaviour Tree, Semantic Decisioning Runtime

Structured Data

Standardised format for providing information about a page and classifying the page content to help search engines understand it. Structured data transforms unstructured web content into machine-readable information.

Also known as: Semantic Markup

Topic Cluster

A content organisation strategy where related articles are structured around a central pillar page, connected through internal links and semantic relationships. In traditional SEO, topic clusters demonstrate topical authority through on-page content depth. In knowledge graph contexts, topic clusters are defined by entity relationships across multiple domains rather than single-site architecture, enabling AI systems to discover authority through graph traversal rather than page-level signals.

Also known as: Semantic network

Triple

The fundamental unit of semantic data in RDF and knowledge graphs, consisting of three components: Subject, Predicate, and Object. A triple expresses a single fact or relationship in machine-readable format. Example: person/adrian-parker → worksFor → #organization. AI agents and GraphRAG systems navigate knowledge graphs by following triples from entity to entity via @id references..

Also known as: RDF Triple

Universal Commerce Protocol

A semantic interchange standard that enables AI agents to discover, evaluate, and transact with commercial entities through machine-readable structured data. The Universal Commerce Protocol defines a common schema-based language for product offers, pricing, availability, and fulfilment metadata — allowing autonomous agents to compare and act on commercial data across vendors and platforms without human mediation.

Also known as: UCP, Universal Commerce Standard

Vector Embeddings

Numerical representations of data that capture semantic meaning, enabling AI systems to understand relationships between concepts. Vector embeddings power semantic search by representing entities as points in high-dimensional space where similar concepts cluster together.

Also known as: Embeddings, Semantic Vectors

Voice Search

Search queries performed through voice commands, requiring natural language processing and conversational AI capabilities for optimal results.

Also known as: Voice Query, Spoken Search