AI resources
Learn the AI terms that matter when building an AI employee for your business. This page explains the basics in plain language, then connects each concept to real SME workflows.
Glossary
Browse the terms you are likely to hear during an AI employee implementation, from basic AI language to workflow design, integrations, privacy, and performance tracking.
AI basics
12 terms
An LLM is an AI model trained to understand and generate language. It can write replies, summarize information, classify text, extract details, and reason through instructions.
Business meaning: For your business, the LLM is the engine behind an AI employee. It needs clear instructions, approved knowledge, and review rules to produce useful work.
A token is a small piece of text that the AI processes. It can be a word, part of a word, punctuation, or spacing.
Business meaning: Tokens affect cost, speed, and how much information can be included. Long documents, chat history, and detailed instructions use more tokens.
A prompt tells the AI what role to take, what information to use, what format to return, and what rules to follow.
Business meaning: Good prompts make outputs more consistent. For AI employees, prompts should include company tone, workflow steps, escalation rules, and examples.
The context window is the amount of text the AI can read in one request, including your instructions, documents, chat history, and the answer it creates.
Business meaning: A bigger context window helps with long documents, but structure still matters. The AI should receive the most relevant information, not every file at once.
RAG lets AI search approved knowledge first, then write an answer based on the retrieved information.
Business meaning: This is useful for FAQs, SOPs, service guides, pricing rules, and internal documents because the AI can answer from your business knowledge.
A vector database stores documents in a way that helps AI find meaning-based matches, not just exact keywords.
Business meaning: It helps an AI employee find the right SOP, policy, product note, or customer history when preparing a reply or summary.
An AI agent can use instructions, tools, and workflow logic to complete a task through multiple steps.
Business meaning: An AI employee is a business-focused AI agent. It should have a narrow role, clear limits, tool access, and human approval for sensitive actions.
A hallucination happens when AI invents details, misreads context, or answers beyond what it actually knows.
Business meaning: Reduce this with approved sources, confidence checks, refusal rules, escalation points, and human review for pricing, promises, complaints, and legal matters.
An API lets one software system send data to another system, such as a website form sending details into a CRM or spreadsheet.
Business meaning: APIs let AI workflows connect to email, WhatsApp tools, forms, CRMs, calendars, spreadsheets, and databases.
Guardrails are instructions, checks, permissions, and review steps that stop AI from doing risky or unwanted actions.
Business meaning: Guardrails help define what the AI can draft, what it can send, what it must ask about, and what it should never decide alone.
Automation moves a task through a defined process without someone manually doing every step.
Business meaning: AI automation works best for repeatable tasks like enquiry summaries, lead follow-ups, document extraction, admin reminders, and weekly reports.
Structured output makes AI return information in a predictable format, such as fields, tables, JSON, or checklist items.
Business meaning: This makes AI easier to connect to spreadsheets, CRMs, forms, reports, task tools, and approval workflows.
Relevant use cases
The goal is not to learn AI vocabulary for its own sake. The goal is to understand enough to choose the right workflow, set the right rules, and avoid risky automation.
Common mistake
Before choosing software, decide what the AI employee should help with, what information it can use, what output it should prepare, and who approves final decisions.
Misconceptions
AI needs your context. Good setup includes source documents, examples, workflow rules, tone, and approval boundaries.
The better starting point is a workflow. Once the workflow is clear, tool selection becomes much easier.
Most SMEs should begin with draft-and-review. Direct sending can come later for low-risk situations after testing.
Relevant, current, and well-structured knowledge is more useful than a large messy folder of outdated files.
Next reading
Once the terms make sense, the next step is to turn the learning into one practical AI employee or automation workflow.
See how a trained assistant is designed around a business workflow, company knowledge, and human approval.
Review the workflow-first process used to audit, scope, build, test, launch, and improve an AI employee.
Explore common first workflows such as enquiry handling, sales follow-up, and admin support.
Need help applying this?
We can help identify the first workflow, structure your business knowledge, define guardrails, and build an AI employee your team can review safely.