AI automation glossary
The terms you will hear in any AI automation project, defined in plain language.
A
- Agentic AI
- Systems where one or more agents plan, act and correct themselves across multiple steps and tools to complete work with limited supervision.
- AI agent
- Software that takes a goal, decides the steps to reach it and uses tools to act, instead of following a fixed script.
- Approval gate
- A point where an agent pauses and asks a named human to approve an action before it happens. Used for anything risky, costly or irreversible. Read the guide
- Audit trail
- A record of every action an agent took, what it read, who approved it and what changed, kept so the work can be reviewed or rolled back. Read the guide
- Automation
- Running a process without manual effort. Traditional automation follows rules; AI automation adds a reasoning step for messy input.
C
- Chatbot
- A conversational interface. A chatbot answers; an agent can also act in your systems.
E
- Embeddings
- Numeric representations of text that let a system find passages by meaning rather than exact words. The basis of most retrieval.
G
- Guardrails
- Rules and checks that constrain what an agent may say or do, such as allowed tools, spend limits, blocked topics and output validation.
H
- Hallucination
- A confident but wrong output from a language model. Managed with grounding in real data, citations, confidence thresholds and human approval.
- Human-in-the-loop
- A design where a person reviews or approves part of an automated process, usually at the point of consequence. Read the guide
L
- LLM
- A large language model: the reasoning and language component an agent uses to interpret input and decide what to do.
M
- MCP (Model Context Protocol)
- An open protocol for connecting models to tools and data sources through a consistent interface, instead of a bespoke integration each time.
O
- Orchestrator
- The component that holds the goal, sequences the agents and tools, manages memory and handles errors across a run.
P
- Prompt
- The instructions and context given to a model. In production systems prompts are versioned and tested like code.
R
- RAG
- Retrieval-augmented generation: finding relevant passages in your own content and giving them to the model so answers are grounded and citable.
- RPA
- Robotic process automation: software that repeats recorded user interface steps. Fast and deterministic, but brittle when screens or inputs change.
T
- Tool use
- An agent calling a defined function or API, for example to search a database, send an email or create a record, instead of only producing text.
W
- Workflow
- A sequence of steps that turns a trigger into an outcome. The unit we map, automate and measure.
Find the first workflow worth automating.
Book a free call and we will tell you honestly whether AI automation helps, and where to start.
We reply within one business day.