Glossary
AI agent terms, in plain language.
Definitions for the AI agent vocabulary that shows up in our engagements, our blog, and the broader 2026 market. Each entry answers one question (what does this term actually mean) without the marketing fluff.
What is the theagency47 glossary?
The theagency47 glossary is a curated set of 18 plain-language definitions of AI agent terminology, organized by category (AI foundations, agent architecture, engagement methodology, compliance, GEO, business). Each entry includes a one-paragraph quick answer (40–120 words), a longer explanation with concrete examples, related terms, and 2–3 FAQ items. Designed to be the reference layer underneath our guides and blog, link to a term and the reader gets the definition without leaving for Wikipedia.
theagency47 · Updated May 2026AI Foundations
- AI Workforce A coordinated team of specialized AI agents (not a single tool) deployed across organizational tiers to handle defined work alongside human employees.
- Context Window The maximum number of tokens an LLM can attend to in a single request, system prompt, conversation history, retrieved documents, and the new user message all count toward it.
- Large Language Model (LLM) A neural network trained on large text corpora that takes text input and produces text output. The underlying engine inside an AI agent, not the agent itself.
- Retrieval-Augmented Generation (RAG) A pattern where an AI agent retrieves relevant documents from a knowledge base before generating a response, grounding its output in specific data rather than only training memory.
Agent Architecture
- Bounded Autonomy The engineering pattern of giving an AI agent specific permissioned actions and explicit escalation rules, letting it act independently within tight, defined boundaries.
- Escalation Rule A configured condition that hands an agent task off to a human when the agent is uncertain, out of scope, or facing high-stakes input.
- Eval Suite A structured set of test cases with known expected outputs that verifies an AI agent's behavior before deployment and after every change.
- Hallucination An AI agent producing output that sounds confident but is factually wrong, typically because the model invented information not present in its inputs.
- Human-in-the-Loop A deployment pattern where every AI agent action is reviewable by a human before downstream effects, used during the first weeks of production.
- Model Context Protocol (MCP) An open standard for connecting AI agents to external data sources, tools, and services through a consistent interface. The 'USB-C for AI integrations.'
- Multi-Agent System (MAS) An AI architecture where multiple specialized agents collaborate on a goal (coordinated through patterns like supervisor, router, handoff, or pipeline) rather than one monolithic agent doing everything.
- Subagent A specialized AI agent invoked by a main agent to handle a bounded task, with its own context window, system prompt, and tool permissions, but no autonomous existence outside its parent.
- System Prompt The instructional text that defines an AI agent's role, voice, boundaries, and decision rules. The single most important customization layer of an agent.
- Tool Use An LLM's ability to call external functions or APIs as part of its response, enabling actions beyond text generation. Also called function calling.
Compliance & Privacy
- AI Guardrails Runtime controls that sit between an application and an AI model, validating inputs and outputs against safety, security, and compliance policies, and blocking or modifying anything that violates them.
- Prompt Injection A class of attacks where malicious instructions are smuggled into an AI agent's input, causing it to ignore its system prompt and follow the attacker's directions instead.
Business
- BDR (Business Development Representative) The salesperson responsible for creating demand that did not exist yet — outbound prospecting into companies that have never heard of you.
- SDR (Sales Development Representative) The salesperson who works inbound leads and books qualified meetings for closers, rather than closing deals themselves.
Next: See our services →
