Back to Learning Center

AI Glossary

60 AI terms in plain English — written for business owners, not engineers.

A

AEO (Answer Engine Optimization)

Optimizing your content so AI assistants and answer engines (ChatGPT, Perplexity, Google AI Overviews) cite your business when customers ask questions — the successor discipline to SEO as more customers ask AI instead of searching.Related: SEO (Search Engine Optimization), Generative AI

Agentic Workflow

A business process where one or more AI agents handle the steps end-to-end — for example intake, qualification, scheduling, and follow-up — with humans reviewing exceptions instead of doing every step manually.Related: AI Agent, Automation, Human in the Loop

AGI (Artificial General Intelligence)

A hypothetical future AI that matches humans across essentially all intellectual work. Today's business AI is 'narrow' — great at specific tasks. If a vendor invokes AGI to sell you software, be skeptical.Related: Artificial Intelligence (AI)

AI Adoption

The process of bringing AI tools into your business — choosing use cases, rolling out tools, training your team, and measuring results. Successful adoption usually starts with one high-impact, low-risk workflow rather than a company-wide overhaul.Related: Change Management, ROI (Return on Investment)

AI Agent

An AI program that doesn't just answer questions but takes actions to complete a goal — looking things up, filling out forms, sending messages, or using other software on your behalf, usually with checkpoints where you stay in control.Example: An agent that watches your inbox, drafts replies to estimate requests, and books jobs on your calendar.Related: Agentic Workflow, Multi-Agent System, LLM (Large Language Model)

AI Readiness

How prepared your business is to get value from AI: whether your data is organized, your processes are documented, and your team is willing and trained. An AI readiness assessment identifies your gaps before you spend money on tools.Related: AI Adoption, Data Hygiene

AI Transcription

Automatically converting speech to text — job-site voice notes, customer calls, meetings. Paired with an LLM, transcripts become summaries, action items, and CRM notes without anyone typing.Related: Voice AI, Natural Language Processing (NLP)

API (Application Programming Interface)

The way two software systems talk to each other. When a tool says it 'integrates via API' with your CRM, it means the two can exchange data automatically without a person copying and pasting.Related: Integration, Webhook

Artificial Intelligence (AI)

Software that performs tasks that normally require human judgment — understanding language, recognizing images, making predictions, or generating content. Modern business AI is mostly powered by machine learning models trained on large amounts of data.Related: Machine Learning (ML), Generative AI, LLM (Large Language Model)

Automation

Making software do repeatable work without a human — sending follow-up texts, moving data between systems, generating invoices. AI-powered automation adds judgment (understanding a message, deciding a next step) to rule-based automation.Related: Workflow Automation, Agentic Workflow, RPA (Robotic Process Automation)

C

Change Management

The people side of adopting new technology: communicating why, training the team, adjusting roles, and handling resistance. Most failed AI projects fail here, not in the technology.Related: AI Adoption

Chatbot

A program that converses with customers by text or voice — on your website, by SMS, or on the phone. Modern AI chatbots understand natural language, answer real questions from your business knowledge, and can book appointments or capture leads 24/7.Example: A website chatbot that answers 'do you service heat pumps?' and books the service call.Related: Conversational AI, Voice AI, LLM (Large Language Model)

Computer Vision

AI that understands images and video — reading meter photos, spotting roof damage from drone footage, verifying job-site safety compliance, or measuring rooms from pictures.Related: Deep Learning, OCR (Optical Character Recognition)

Context Window

How much text an AI model can 'see' at once — its short-term memory. If a document exceeds the context window, the model can't consider all of it at the same time, which is why long jobs get split up or use retrieval techniques.Related: LLM (Large Language Model), Token, RAG (Retrieval-Augmented Generation)

Conversational AI

AI systems designed to hold natural back-and-forth conversations — chatbots, voice assistants, and AI phone agents. Distinguished from simple scripted bots by their ability to understand intent and handle unexpected phrasing.Related: Chatbot, Voice AI, Natural Language Processing (NLP)

Copilot

A general label for AI assistants embedded inside software you already use (email, spreadsheets, CRMs) that draft, summarize, and suggest while you keep control. Microsoft brands its assistants this way, but the pattern is everywhere.Related: Generative AI, LLM (Large Language Model)

CRM (Customer Relationship Management)

The software system where you track customers, leads, jobs, and communication history. Your CRM is usually the most valuable data source for AI: lead scoring, follow-up automation, and customer service AI all plug into it.Related: Integration, Lead Scoring

D

Data Hygiene

Keeping your business data accurate, consistent, and organized — no duplicate customers, current phone numbers, jobs categorized the same way. AI tools amplify whatever data you feed them, so clean data is a prerequisite for useful results.Related: AI Readiness, CRM (Customer Relationship Management)

Deep Learning

The branch of machine learning behind most modern AI, using layered 'neural networks' loosely inspired by the brain. It's what enables computers to understand speech, images, and language well enough to be useful in business.Related: Machine Learning (ML), Neural Network

E

Embedding

A way of turning text (or images) into lists of numbers that capture meaning, so a computer can measure how similar two things are. Embeddings power semantic search — finding 'AC not cooling' when a customer typed 'air conditioner blows warm'.Related: Semantic Search, Vector Database, RAG (Retrieval-Augmented Generation)

F

Fine-Tuning

Additional training that adapts a general AI model to your specific domain or style using your own examples. Most small businesses don't need fine-tuning — good prompts and retrieval (RAG) usually get you there cheaper.Related: LLM (Large Language Model), RAG (Retrieval-Augmented Generation), Prompt Engineering

G

Generative AI

AI that creates new content — text, images, audio, video, or code — rather than just analyzing existing data. ChatGPT, Claude, Gemini, and Midjourney are generative AI. It's the technology behind AI-written estimates, marketing copy, and job photos cleanup.Related: LLM (Large Language Model), Artificial Intelligence (AI), Hallucination

GPT

Short for 'Generative Pre-trained Transformer' — the family of large language models from OpenAI that powers ChatGPT. Often used loosely to mean any AI chat model.Related: LLM (Large Language Model), Generative AI

Grounding

Connecting an AI's answers to real, verified sources — your price book, your service policies, current data — instead of letting it rely on its general training. Grounded AI cites where answers came from and makes fewer things up.Related: RAG (Retrieval-Augmented Generation), Hallucination, Knowledge Base

H

Hallucination

When an AI confidently states something false — a made-up price, a nonexistent policy, a wrong spec. The risk is managed by grounding the AI in your real data, keeping humans in the loop for consequential decisions, and asking for sources.Related: Grounding, Human in the Loop

Human in the Loop

A design where AI does the heavy lifting but a person reviews or approves before anything consequential happens — sending a quote, issuing a refund, publishing content. The standard safety pattern for business AI.Related: AI Agent, Hallucination

I

Inference

The act of an AI model actually producing an answer (as opposed to being trained). When pricing says 'per inference' or 'per generation', you're paying each time the model runs.Related: Token, LLM (Large Language Model)

Integration

A connection that lets two software tools share data automatically — your booking system updating your CRM, your AI receptionist logging calls. When evaluating AI tools, 'what does it integrate with?' is often the deciding question.Related: API (Application Programming Interface), Webhook, Workflow Automation

IoT (Internet of Things)

Everyday equipment — thermostats, water heaters, fleet vehicles — connected to the internet and reporting data. IoT sensors supply the data that makes predictive maintenance and smart dispatching possible.Related: Predictive Maintenance

K

Knowledge Base

The organized collection of your business information — services, prices, policies, FAQs, procedures — that an AI assistant draws on to answer questions accurately. Building one is usually step one of a customer-facing AI project.Related: Grounding, RAG (Retrieval-Augmented Generation)

L

Lead Scoring

Using AI to rank incoming leads by how likely they are to become paying customers, based on patterns in your historical data — so your team calls the hottest leads first.Related: CRM (Customer Relationship Management), Predictive Analytics

LLM (Large Language Model)

The engine behind modern AI chat tools — a model trained on enormous amounts of text that can read, write, summarize, translate, and reason in plain language. ChatGPT, Claude, and Gemini are all built on LLMs.Related: GPT, Generative AI, Token, Context Window

M

Machine Learning (ML)

Software that learns patterns from data instead of following hand-written rules. It's how AI systems predict which equipment will fail, which leads will convert, or what a customer message means.Related: Artificial Intelligence (AI), Deep Learning, Predictive Analytics

MCP (Model Context Protocol)

An open standard that lets AI assistants securely connect to your business tools and data — calendars, CRMs, databases — so the AI can look things up and take actions in the systems you already use.Related: AI Agent, API (Application Programming Interface), Integration

Multi-Agent System

Several AI agents working together, each with a specialty — one qualifies the lead, another checks the schedule, a third drafts the confirmation. Useful when one agent handling everything would be unreliable.Related: AI Agent, Agentic Workflow, Orchestration

N

Natural Language Processing (NLP)

The field of AI focused on understanding and generating human language. It's why software can now read a rambling customer email and extract the address, the problem, and the urgency.Related: LLM (Large Language Model), Conversational AI, Sentiment Analysis

Neural Network

The layered mathematical structure that modern AI models are built from, loosely inspired by neurons in the brain. You don't need to understand the math — just know that 'deep learning' means big neural networks.Related: Deep Learning, Machine Learning (ML)

No-Code / Low-Code

Tools that let non-programmers build software and automations by dragging, dropping, and configuring instead of writing code. Platforms like Make, n8n, and Zapier let small businesses build AI workflows without hiring developers.Related: Workflow Automation, Vibe Coding

O

OCR (Optical Character Recognition)

Technology that reads text from images and scans — turning a photo of a receipt, invoice, or serial-number plate into data your software can use.Related: Automation

Orchestration

Coordinating multiple AI models, agents, or automation steps into one reliable workflow — deciding what runs when, what happens on failure, and where humans check in.Related: Multi-Agent System, Workflow Automation

P

Predictive Analytics

Using historical data to forecast what's likely to happen — which HVAC units will fail this summer, which customers are likely to churn, how many calls Tuesday will bring — so you can act before it happens.Related: Machine Learning (ML), Predictive Maintenance

Predictive Maintenance

Using sensor data and AI to spot equipment problems weeks before failure, so service happens on schedule instead of as an emergency. A flagship AI use case in HVAC, plumbing, and facilities.Example: A monitored compressor showing the failure pattern gets a maintenance visit before the July heat wave.Related: Predictive Analytics, IoT (Internet of Things)

Prompt

The instruction you give an AI model. The quality of what you get out depends heavily on what you put in: context about your business, the format you want, examples, and constraints.Related: Prompt Engineering, System Prompt

Prompt Engineering

The craft of writing instructions that get consistently good results from AI — providing role, context, examples, output format, and guardrails. The highest-leverage AI skill a business owner can learn, because it costs nothing.Related: Prompt, System Prompt

R

RAG (Retrieval-Augmented Generation)

A technique where the AI first looks up relevant information from your documents, then writes its answer based on what it found. It's how an AI assistant answers from YOUR price book and policies instead of guessing.Related: Grounding, Knowledge Base, Embedding, Vector Database

ROI (Return on Investment)

What you get back for what you spend. For AI projects, measure it in hours saved, leads captured, jobs booked, and errors avoided — and compare against subscription plus setup costs. If you can't name the metric, don't buy the tool yet.Related: AI Adoption

RPA (Robotic Process Automation)

Software 'robots' that mimic human clicks and keystrokes to automate repetitive computer tasks in systems that lack integrations. Older than modern AI, and increasingly combined with it.Related: Automation, Workflow Automation

S

Sentiment Analysis

AI reading text or listening to calls to gauge emotion — flagging an angry customer email for immediate attention or scoring how calls are going across the team.Related: Natural Language Processing (NLP)

SEO (Search Engine Optimization)

Making your website more likely to show up when customers search Google. AI now assists with SEO (content, keywords, technical audits) and is also changing it — see AEO.Related: AEO (Answer Engine Optimization)

Structured Data (Schema Markup)

Machine-readable labels added to web pages that tell search engines and AI exactly what the page contains — your services, prices, FAQs, reviews. It makes your business easier for answer engines to quote correctly.Related: SEO (Search Engine Optimization), AEO (Answer Engine Optimization)

System Prompt

The standing instructions an AI assistant always follows — its role, tone, rules, and boundaries — set once by whoever configures it, separate from what users type. Your AI receptionist's personality and policies live in its system prompt.Related: Prompt, Prompt Engineering

T

Token

The chunks (roughly word-fragments) that AI models read and write text in — about 750 words per 1,000 tokens. AI pricing is usually per token, which is why long documents cost more to process.Related: LLM (Large Language Model), Context Window, Inference

Training Data

The examples an AI model learned from. A model's strengths, blind spots, and cutoff date all come from its training data — which is why grounding it in your current business data matters.Related: Machine Learning (ML), Fine-Tuning, Grounding

V

Vector Database

A database built to store embeddings and find 'things similar in meaning' fast. It's the storage layer behind semantic search and RAG — usually invisible to you, but you'll see the term in tool descriptions.Related: Embedding, RAG (Retrieval-Augmented Generation), Semantic Search

Vibe Coding

Building software by describing what you want to an AI in plain English and iterating on what it generates, rather than writing code by hand. Tools like Cursor, Lovable, and Aura make it practical for non-developers to build real internal tools.Related: No-Code / Low-Code, Generative AI

Voice AI

AI that speaks and listens — answering your business phone, qualifying callers, booking appointments, and escalating to a human when needed. One of the fastest-payback AI tools for service businesses that miss calls.Example: An AI receptionist that answers after-hours calls and books the morning's first appointment.Related: Conversational AI, Chatbot, AI Transcription

W

Webhook

A simple 'push notification between apps': when something happens in one tool (new lead, completed job), it instantly sends the data to another tool. The plumbing behind many no-code automations.Related: API (Application Programming Interface), Integration, Workflow Automation

Workflow Automation

Connecting your tools so multi-step processes run themselves: lead comes in → CRM record created → welcome text sent → task assigned. Platforms like Make, n8n, and Zapier are the usual glue, with AI adding judgment steps.Related: Automation, No-Code / Low-Code, Integration

Z

Zero-Shot / Few-Shot

How much example-giving an AI needs: 'zero-shot' means it performs a task from instructions alone; 'few-shot' means you show it a handful of examples in the prompt first. Adding two or three examples is often the cheapest way to improve results.Related: Prompt Engineering, Prompt

Know the words. Now use them.

The fastest way from vocabulary to results: follow a guided learning path or grab ready-made prompts.

Ask JustStart AI

Search the whole site

Real pages and resources only. No generated answers.

Popular searches

Search titles, summaries, article headings, tags, glossary terms, prompts, learning paths, FAQs, videos, courses, tools, and public routes.