Glossary

The AI tools glossary for businesses

Plain-English definitions of the AI tools and business terms that come up as you build.

AI agent

An AI system that can take actions on its own to reach a goal - not just answer a question, but read, decide, and do multiple steps.

Prompt

The instruction you give an AI tool. The clearer you describe the outcome you want, the better the result you get back.

Prompt engineering

The skill of writing instructions that get useful results from AI tools - mostly a matter of clarity about the outcome, not technical tricks.

Workflow automation

Connecting your apps so a repetitive, multi-step task runs itself - often with AI steps handling the parts that need a little judgement.

No-code

Building software without writing code, using visual tools. AI has pushed this further - you can now describe what you want in plain English.

LLM (large language model)

The kind of AI model behind most current tools - trained on huge amounts of text to understand and generate language, code, and structured answers.

Prototype

A quick, rough version of an idea built to test it. AI tools make prototypes cheap and fast, so you can check demand before you build for real.

Productized service

A repeatable deliverable sold at a fixed scope and price, instead of scoping every project from scratch - easier to price, deliver, and scale.

MCP (Model Context Protocol)

An open standard that lets AI tools connect to your other apps and data sources in a consistent way, so an agent can act across the tools you use.

API

A way for software to talk to other software. AI tools use APIs to connect to your apps; automations wire APIs together so work flows between them.

Token

The small chunk of text an AI model reads and writes in - roughly a word or part of a word. Most AI pricing and limits are counted in tokens.

Context window

The amount of text an AI model can hold in mind at once. Go past it and the model starts forgetting the earliest parts of the conversation.

Hallucination

When an AI states something false with full confidence. It is a known limit of current models, which is why you check anything that matters.

Retrieval (RAG)

Giving an AI access to your own documents so it answers from your real information instead of guessing - the reliable way to build an AI over your content.

Fine-tuning

Further training an existing AI model on your own examples so it adapts to your style or task. Often overkill - retrieval or a good prompt does the job cheaper.

Model

The trained AI engine that does the actual thinking. Different models trade off speed, cost, and capability - you pick one to suit the task.

Inference

The moment an AI model actually runs to produce an answer. Every question you ask triggers inference, and that is usually what you pay for.

Training data

The huge collection of text, images, or examples a model learned from. Its strengths and blind spots trace back to what it was trained on.

Machine learning

Software that learns patterns from examples instead of being programmed with fixed rules. It is the broad field that modern AI tools grew out of.

Neural network

The layered structure, loosely inspired by the brain, that most AI models are built on. You do not need to understand it to use the tools well.

GPT

A family of large language models that popularized modern AI chat. In everyday use, people say GPT to mean the general kind of AI that generates text.

Multimodal

An AI model that handles more than one kind of input - text plus images, audio, or video - so you can show it a photo or file, not just type.

Vision model

An AI that can look at images and understand them - reading a receipt, describing a photo, or checking a product picture against a rule.

Text-to-speech

AI that turns written words into natural-sounding spoken audio - used for voiceovers, phone systems, and accessibility.

Speech-to-text

AI that turns spoken audio into written text - the engine behind live captions, meeting notes, and voice transcription.

Embedding

A way of turning text into numbers that capture its meaning, so software can find related content by similarity rather than exact wording.

Vector database

A store built to search by meaning rather than keywords, using embeddings. It is the memory layer behind most retrieval-based AI tools.

Low-code

Building software mostly through visual tools, with the option to drop into a little code when you need it - a step beyond no-code.

Workflow

The set steps a task moves through from start to finish. Mapping your workflow is the first move before you automate any part of it.

Automation

Setting up software to do a repetitive task for you, so it runs on its own without you touching it each time.

Webhook

An automatic message one app sends another the instant something happens - the trigger that kicks off many automations in real time.

Integration

A connection between two tools that lets them share data and work together, instead of you copying information across by hand.

Chatbot

A tool that talks with people in plain language to answer questions or take actions - now far more capable thanks to modern AI models.

Copilot

An AI assistant that works alongside you inside a tool - suggesting, drafting, and speeding you up, while you stay in control of the result.

Agentic

Describing AI that acts on its own across several steps to reach a goal, rather than just replying once. The trait that makes an AI agent an agent.

System prompt

The standing set of instructions that shapes how an AI behaves across a whole tool - its role, tone, and rules - separate from what a user types.

Temperature

A setting that controls how varied an AI's answers are. Lower gives steady, predictable replies; higher gives more creative, less predictable ones.

Guardrails

The limits you put around an AI so it stays on task and out of trouble - what it can access, what it must not say, and where a human signs off.

Vibe coding

Building software by describing what you want in plain English and letting an AI write the code, instead of writing it yourself line by line.

AI wrapper

A product built on top of an existing AI model, adding a focused interface and workflow around it. Many useful tools are wrappers, and that is fine.

Deployment

Putting your software live so real people can use it, rather than it only running on your own machine. Modern tools make deploying a few clicks.

Frontend

The part of an app people see and click - the screens, buttons, and layout. It is the surface that sits on top of the behind-the-scenes logic.

Backend

The behind-the-scenes part of an app - the logic, data, and connections that make it work, hidden from the person using it.

Database

The organized store where an app keeps its information so it can be saved, searched, and updated - the memory behind almost every tool.

SaaS

Software you subscribe to and use through a browser instead of installing it - the model behind most business tools you pay a monthly fee for.