AI glossary for business owners

Jump to: AI basics · How models work · Prompts · Memory and data · Agents and automation · Quality and risk · Business uses · AI search

AI basics

Artificial intelligence (AI)

Software that performs tasks that normally need human judgment, such as understanding language, recognizing images or making predictions.

Generative AI

AI that creates new content (text, images, audio, video or code) instead of only classifying or predicting.

Large language model (LLM)

An AI model trained on huge amounts of text to predict the next word. LLMs power ChatGPT, Claude, Gemini and most business chat and writing tools.

Foundation model

A large general-purpose model, such as GPT, Claude, Gemini or Llama, that other products are built on top of.

Multimodal model

A model that can take in or produce more than one kind of input, for example reading a photo of an invoice and answering questions about it.

Reasoning model

A model that works through a problem step by step before answering. Slower and more expensive per answer, but better at math, planning and multi-step tasks.

Small language model (SLM)

A compact model that runs cheaply or on a single device. Good enough for narrow tasks like sorting emails or extracting fields.

Open-weight model

A model whose trained weights are published, so a business can run it on its own servers, such as Llama, Mistral or Qwen.

Chatbot

Software that holds a conversation. Older chatbots followed fixed scripts; modern ones use an LLM to understand free-form questions.

How models work

Token

The unit an LLM reads and writes, roughly three quarters of an English word. AI providers price usage per million tokens.

Context window

How much text a model can consider at once, measured in tokens. It includes the instructions, the documents you pass in and the conversation so far.

Training data

The text, images and other material a model learned from.

Knowledge cutoff

The date after which a model has no training data. Anything newer has to come from search, documents or tools.

Inference

Running a trained model to get an answer. Each request is an inference, and it is what you pay for when you use an AI API.

Parameters

The internal numbers a model adjusts during training. More parameters usually means more capability and higher running cost.

Temperature

A setting that controls how random a model's output is. Low values give consistent answers; high values give more varied, creative ones.

Latency

How long a model takes to respond. It matters for live uses such as replying to a new lead or answering a phone call.

API

Application programming interface: how software talks to an AI model or another system directly, without a person using a chat window.

Prompts

Prompt

The instructions and information given to a model for a task.

System prompt

Standing instructions that set a model's role, rules and tone for every conversation, usually hidden from the end user.

Prompt engineering

Writing and testing prompts so a model produces reliable, useful output for a specific job.

Few-shot prompting

Including a few worked examples in a prompt so the model copies the pattern.

Structured output

Asking a model to answer in a fixed format such as JSON so other software can read the result without cleanup.

Memory and data

Embedding

A list of numbers that represents the meaning of a piece of text, image or audio. Content with similar meaning gets similar numbers, which lets software search by meaning instead of exact words.

Vector database

A database built to store embeddings and quickly find the closest matches. Examples include Pinecone, Weaviate, Qdrant and pgvector for Postgres.

Vector memory

Long-term memory for an AI assistant or agent built on a vector database. Past conversations, notes or documents are stored as embeddings, and the most relevant ones are pulled back in when needed.

Semantic search

Search that matches meaning rather than keywords, so "my AC is blowing warm air" finds a page about air conditioner repair.

Retrieval-augmented generation (RAG)

A setup where the system first searches your own documents for relevant passages, then gives them to the model to answer from. It keeps answers grounded in your data without retraining the model.

Chunking

Splitting long documents into smaller passages before creating embeddings, so search can return the specific section that answers a question.

Knowledge base

The collection of documents, FAQs, policies and records an AI system is allowed to draw answers from.

Knowledge graph

A map of entities (people, businesses, products, places) and how they relate. Search engines use them to understand who a business is.

Fine-tuning

Further training an existing model on your own examples so it adopts a style or learns a narrow task. Often unnecessary when good prompts and RAG do the job.

Synthetic data

Data generated by AI to train or test other systems when real data is scarce or sensitive.

Agents and automation

AI agent

An AI system that takes actions toward a goal instead of only answering: it looks things up, uses tools, updates a CRM or drafts a message, then decides the next step.

Agentic workflow

A process where one or more AI agents handle several steps on their own, with checkpoints where a person approves anything important.

Tool use (function calling)

A model's ability to call outside software, such as a calendar, CRM or search engine, by producing a structured request that the software runs.

Model Context Protocol (MCP)

An open standard, introduced by Anthropic in 2024, for connecting AI assistants to tools and data sources through a common interface, so one connector works across many AI apps.

Multi-agent system

Several specialized agents working together, for example one that researches, one that writes and one that checks the work.

Orchestration

The logic that decides which model, agent or tool runs at each step of a workflow, and in what order.

Human-in-the-loop

A design where a person reviews or approves AI output before it goes out or changes anything that matters.

Workflow automation

Connecting apps so routine steps run on their own, using platforms like Zapier, Make.com or n8n, increasingly with AI steps inside.

Webhook

An instant message one app sends another when something happens, such as a new form submission. It is how speed-to-lead systems react within seconds.

AI voice agent

An AI that answers or makes phone calls in natural speech, qualifies callers and books appointments.

Quality and risk

Hallucination

When a model states something false with confidence. Grounding answers in real data and adding checks reduces it.

Guardrails

Rules and filters that keep an AI system on task and stop unsafe, off-brand or out-of-scope output.

Evals

Repeatable tests that measure how well an AI system performs a task, used to compare prompts or models before going live.

Prompt injection

An attack where hidden or malicious instructions in an email, web page or document try to make an AI system ignore its rules.

PII (personally identifiable information)

Names, phone numbers, addresses, and health or financial details that identify a person. The same privacy rules apply whether a person or an AI system handles them.

AI governance

The policies a business sets for which AI tools staff may use, what data may go into them and who reviews the output.

Business uses

Speed to lead

How fast a business responds to a new inquiry. A Harvard Business Review study of 2,241 companies found firms that replied within an hour were far more likely to qualify the lead than those that waited longer.

Lead reactivation

Re-engaging old or cold leads in a CRM with personalized outreach, often AI-assisted, to book appointments from contacts already paid for.

AI receptionist

An AI system that answers calls, texts or chats around the clock, captures details and books appointments.

Copilot

An AI assistant built into software people already use, such as email, documents or a CRM, that drafts and suggests while a person stays in control.

AI automation ROI

The return on an AI project: hours saved, revenue recovered or errors avoided, compared with the cost to build and run it.

AI search

AI search visibility

How often, and how favorably, AI assistants mention or recommend a business when people ask about its products or services.

Generative engine optimization (GEO)

Improving how a business is represented in answers from AI systems such as ChatGPT, Gemini, Perplexity and Google AI Overviews.

Answer engine optimization (AEO)

Structuring content so it can be lifted directly as an answer: by AI assistants, featured snippets and voice assistants. Often used interchangeably with GEO.

Google AI Overviews

AI-generated summaries that appear above regular results in Google Search, citing a handful of sources.

AI Mode

Google's conversational search experience, which answers follow-up questions using AI and links to sources.

Citation

A link or source an AI answer credits. Being cited sends traffic and signals trust.

Share of voice

The percentage of relevant AI answers that mention your business compared with competitors.

Structured data (schema markup)

Code on a page, usually JSON-LD, that labels facts like business name, address, services, hours and reviews so machines read them without guessing.

LocalBusiness schema

The schema.org type for physical or service-area businesses, with subtypes such as Plumber, HomeAndConstructionBusiness or MedicalSpa.

llms.txt

A proposed plain-text file at the root of a website that summarizes the site for large language models, similar in spirit to robots.txt. Server-log research in 2026 found most llms.txt files are rarely requested by AI bots, so treat it as a low-cost extra, not a ranking lever.

AI crawlers

Bots that collect web content for AI systems, such as GPTBot and OAI-SearchBot (OpenAI), ClaudeBot (Anthropic), PerplexityBot and Google-Extended.

robots.txt

A file that tells crawlers which parts of a site they may visit. Blocking AI crawlers here can remove a business from AI answers.

Entity

A thing search engines and AI systems recognize as distinct: a business, person or place. Consistent name, address and details across the web strengthen it.

NAP consistency

Name, address and phone number matching exactly across your website, Google profile and directories.

Grounding

When an AI assistant looks up live web or map data to support its answer instead of relying only on what it learned in training.

Query fan-out

When an AI search tool splits one question into several related searches behind the scenes and combines the results into one answer.

Zero-click search

A search answered on the results page or in an AI answer, so the user never visits a website.

Next: the AI visibility checklist, or see what speed-to-lead automation looks like in practice.