AI Automation Glossary: 20 Terms Explained Simply
Twenty terms you will hear in every AI automation conversation — defined without the sales pitch.
Plain-language definitions for the terms that come up in every automation project. If a provider cannot explain these without jargon, ask harder questions.
AI agent
A system that can take actions, not just answer questions. It uses a model plus tools such as a CRM, calendar or database, and usually has permissions and a human approval step.
AI automation
Using AI to handle a workflow that previously needed a person: reading a message, deciding what it means, updating a system and sending the next step.
API
A documented way for two software systems to exchange data. Integrations are built on APIs, which is why access and rate limits matter.
Chatbot
A conversational interface on a website or messaging channel. It can answer questions, capture leads and hand over to a person.
CRM
Customer relationship management software that stores leads, contacts and deal history. HubSpot, Salesforce and GoHighLevel are common examples.
Data enrichment
Adding missing details to a record — company size, industry, location — so routing and personalisation rules can work.
Guardrails
Limits placed on an AI system: allowed actions, approved data, required confirmations and escalation rules when confidence is low.
Human-in-the-loop
A workflow where a person reviews or approves an action before it becomes final, such as sending a quote or issuing a refund.
Integration
A connection between two tools so data moves automatically instead of being copied by hand.
Lead qualification
The process of deciding whether an enquiry fits your service, budget and timeline, and routing it to the right person.
LLM
Large language model — the model behind tools like ChatGPT and Claude. It predicts text and can be given tools and instructions.
No-code automation
Building workflows in tools like Make, Zapier or n8n without writing a full application. Fast for simple logic, fragile for complex edge cases.
OCR
Optical character recognition: converting scanned documents and PDFs into text that software can read.
Prompt
The instruction given to a language model. Good prompts define the role, the input, the output format and what to do when information is missing.
RAG
Retrieval-augmented generation: the model retrieves relevant documents first, then answers using them. It reduces invented answers and keeps responses current.
Rate limit
A cap on how many requests a system accepts in a period. It protects APIs and is a common cause of automation failures if ignored.
Schema markup
Structured data added to a page so search engines and AI systems can understand entities, reviews, FAQs and prices.
Webhook
An automatic message one system sends to another when an event happens, such as a form submission or a payment.
Workflow
A defined sequence of steps that turns an input into an outcome, including the exceptions and who handles them.
Workflow orchestration
Coordinating multiple workflows and systems so they run in the right order, with retries and visibility when something fails.
Want these applied to your workflow? Book a free consultation and we will map it in plain English.