GLOSSARY
AI and automation glossary
Plain definitions of the terms that come up when you build agents and automation. No jargon for its own sake.
AGENTS
AI agent
A system that uses a model to reason over a task, call tools, and take steps toward a goal rather than only answering a single question.
Always-on agent
An agent that runs on its own computer and keeps working on a goal between conversations, instead of only replying when you ask.
Agentic workflow
A process where a model decides the next step, mixed with fixed rules, rather than following a path you defined in advance.
Multi-agent
Several agents working together, each with a role, passing work between them.
Computer use
An agent controlling a browser or desktop interface the way a person would, which lets it work with tools that have no API.
Orchestration
Coordinating the steps, tools, and agents in a workflow so they run in the right order and hand off correctly.
KNOWLEDGE AND MODELS
LLM
Large language model. The model behind chat assistants and many agents.
RAG
Retrieval-augmented generation. The system retrieves relevant material first, then the model answers from it, with a source.
Embeddings
Numeric representations of text that let a system find pieces of content that mean something similar.
Vector database
A store built to search those representations quickly. Common options include pgvector, Pinecone, Qdrant, and Weaviate.
Fine-tuning
Training a model further so it behaves a certain way. It changes behavior, not knowledge, and is expensive to keep current.
Hallucination
When a model states something false with confidence. Grounding in retrieved sources is the usual fix.
CONTROL AND SAFETY
Human-in-the-loop
A step where a person reviews or approves before the system continues.
Approval boundary
The line that decides which actions need a person: typically anything that sends, spends, or deletes.
Audit trail
A record of what the system did, with what input, and what changed, so actions can be traced.
Sandboxing
Running an agent in an isolated environment so a mistake stays contained.
Idempotency
Making sure the same request run twice does not create two results. This is what stops duplicate records.
Observability
Seeing what a system is doing in production: logs, traces, and metrics, so problems surface before users notice.
CONNECTING SYSTEMS
API
A defined way for one system to talk to another.
Webhook
A message one system sends to another when something happens, instead of the second system asking repeatedly.
MCP
Model Context Protocol. A standard that lets an agent call tools and read data through one reusable connection.
A2A
Agent to agent. A standard for agents to hand work to each other.
TOOLS AND PLATFORMS
n8n
A source-available workflow automation platform you can self-host or run in the cloud, with strong support for custom logic and AI.
Make
A visual automation platform with strong branching and error handling, aimed at non-developers.
Zapier
The widest app coverage and the easiest start, and the most expensive at high volume.
GoHighLevel
An all-in-one CRM and marketing platform for agencies and service businesses.
LangChain
A framework for building AI applications and agents with a standard tool-calling architecture.
LangGraph
A runtime for stateful agents with durable execution and human-in-the-loop control.
Voice agent
An agent that handles calls: it listens, understands, speaks, and can take actions like booking.
Not sure how a term applies to your work? See the tools I work with or the guides.