Contact, if you are interested in this website / domain name / Sponsorship / Advertisement / Partnership: web.works/contact

AI glossary for business owners

The 30 terms you will meet when buying or using AI tools, explained without jargon.

Agent

An AI system that can take a series of actions toward a goal, such as reading an email, checking a calendar and booking a meeting, instead of only replying with text.

API

Application programming interface. A way for software to talk to other software. Automation tools use APIs to move data between your apps.

Automation

Work that runs on its own when a trigger happens, for example: a new form entry creates a CRM contact and sends a welcome email.

Chatbot

A program that answers questions in a chat window. Modern chatbots use large language models and your help content to answer in natural language.

Context window

How much text a model can consider at once, measured in tokens. A bigger window lets you work with longer documents in one go.

Copilot

An AI assistant built into a tool you already use, such as a word processor, spreadsheet or code editor.

Embedding

A list of numbers that represents the meaning of text. Embeddings let software find passages that mean the same thing even with different words.

Fine-tuning

Training an existing model further on your own examples so it follows your style or task more closely. Most small businesses do not need it; good prompts and retrieval usually suffice.

Generative AI

AI that creates new content, such as text, images, audio, video or code, rather than only classifying or predicting.

Grounding

Making an AI answer from specific trusted sources, such as your policies or product data, to reduce made-up answers.

Guardrails

Rules and checks that keep AI output safe and on-topic, such as blocked topics, required disclaimers or human approval for refunds.

Hallucination

When a model states something false with confidence. The fix is grounding, asking for sources and human review of anything that matters.

Human in the loop

A step where a person reviews or approves AI work before it goes out. Recommended for anything customer-facing or financial.

Inference

Running a trained model to get an answer. Tool pricing is often based on inference usage.

Integration

A connection between two apps so they share data automatically.

Large language model (LLM)

A model trained on huge amounts of text to predict the next words. It powers assistants such as ChatGPT, Claude and Gemini.

MCP (Model Context Protocol)

An open standard that lets AI assistants connect to tools and data sources in a consistent way.

Multimodal

A model that understands or produces more than one type of input, for example text plus images or audio.

No-code

Building workflows or apps with visual editors instead of programming.

Open-source model

A model whose weights are published so anyone can run or adapt it, often on their own servers for privacy or cost reasons.

Prompt

The instruction you give an AI tool. Good prompts state the role, context, task and output format.

Prompt library

A shared set of tested prompts so a whole team gets consistent results.

RAG (retrieval-augmented generation)

A method that searches your documents first and gives the relevant passages to the model, so its answer is based on your information.

ROI

Return on investment: the gain from an investment compared with its cost. For AI, mostly hours saved times the value of an hour, minus tool and setup costs.

Seat

One user license. Many AI tools price per seat per month.

System prompt

Standing instructions that shape every response of an AI tool, such as tone, rules and what it must never do.

Token

A chunk of text, roughly three-quarters of a word in English. Models read, write and are often priced in tokens.

Trigger

The event that starts an automation, such as a new email, a form submission or a scheduled time.

Vector database

A database that stores embeddings so software can quickly find text with similar meaning. Common in RAG systems.

Workflow

A repeatable sequence of steps to get a job done. AI works best when it is placed inside a clear workflow rather than used ad hoc.

Ready to apply them? Start with how to build an AI stack or browse the tools directory.

The Friday Stack

One email a week: three AI tools worth testing, one workflow you can copy, and what changed in pricing. Five minutes to read. No spam, unsubscribe in one click.