Agent
An AI system that can take actions (not just generate text) within defined rules, constraints, and permissions.
Glossary
Glossary of Terms
A plain-English guide to the concepts behind context-first AI — without the jargon, and focused on how they show up in real business workflows.
Why this page exists
A plain-English guide to context-first AI
AI conversations are full of new terms — agents, RAG, skills, orchestration, automation.
Most explanations assume technical knowledge and miss the business meaning.
This glossary exists to ground those terms in how AI actually works inside a business — simply, clearly, and without jargon.
These definitions reflect how VisionList uses each concept inside a context-first operating model.
Definitions
How VisionList uses these terms
The entries below explain how each concept is applied inside a context-first operating model. Use them when aligning AI teams, onboarding people, or publishing your V-Wallet.
Agentic Harness
The controls that keep AI agents aligned — including context, boundaries, escalation rules, and governance — so autonomy does not become chaos.
Agentic Interface
The layer between human intent and AI execution. It translates goals, constraints, and priorities into action without constant supervision.
Agent-Ready Definition (ARD)
A structured instruction set that defines what an AI agent is allowed to do, decide, escalate, or defer — ensuring control and repeatability.
Automation
The execution of tasks by AI or software systems. Automation works best after intent and constraints are clearly defined.
Context
The explicit information AI needs to reason correctly — including goals, constraints, priorities, assumptions, and decision boundaries. Context is not data volume or long prompts.
Context-First Operating Model
An approach where business intent is defined before automation begins, keeping AI downstream of human judgment.
Drift
When AI outputs slowly lose alignment with business intent due to missing, outdated, or implicit context.
Execution Metadata
Information about how decisions are made over time — including priorities, learnings, reviews, and governance — used to keep AI consistent.
RAG (Retrieval-Augmented Generation)
A method for retrieving relevant information. RAG supports context but does not replace the need for defined intent and constraints.
Recursive Language Models (RLM)
RLMs improve how context is reused, but not how intent is defined. They reduce retrieval noise, yet still rely on pre-existing assumptions about goals, priorities, and decision logic — which must be managed elsewhere.
Sandbox
A workspace for defining, refining, and testing business context before scaling automation or agents.
Unified Context Layer (UCL)
A structured, machine-readable representation of how a business thinks and operates — the environment AI systems run within.
The VisionList UCL is composed of five documents:
- VDD (Vision Definition Document) — Defines the opportunity, business model, and desired outcomes.
- XDD (Extended Definition Document) — Specifies offers, requirements, constraints, and campaign intent.
- SCD (System Context Document) — Captures processes, systems, dependencies, and operating assumptions.
- EMD (Execution Metadata Document) — Records priorities, decisions, learnings, and governance rules.
- ARD (Agent-Ready Definition) — Defines permissions and boundaries for AI agents.
V-Wallet
A secure, shareable link containing your AI-ready business context — allowing AI tools and agents to operate without repeated explanation.
VisionList
A context-first platform designed to help teams define, maintain, and share the business context AI needs to operate reliably.
VisionList and Claude Cowork
Claude Cowork is an agent execution environment that enables long-horizon reasoning and tool use. VisionList is an upstream business context operating layer that defines intent, rules, constraints, and escalation logic for agents. VisionList does not replace agent platforms — it governs them.
Keep going
When these terms are documented inside VisionList, AI systems gain the clarity they need to operate confidently. Continue exploring how the platform helps you install and maintain that context.