AI and automation glossary

A

Agentic workflow

A process in which an AI system is given a goal and several steps to carry out, such as researching, drafting and filing, with checks along the way rather than a single answer.

AI agent

Software that uses an AI model to decide on and carry out a series of actions toward a goal, such as looking something up, then writing and sending a reply for approval.

AI governance

The rules, roles and checks a business sets for how AI is chosen, used and reviewed, including who is responsible and what always needs a person's approval.

Answer engine optimization (AEO)

Making a business's information clear, accurate and well structured so AI assistants and answer engines describe it correctly when someone asks.

API (application programming interface)

A defined way for one piece of software to ask another for information or to carry out an action, such as a booking tool adding an event to a calendar.

Artificial intelligence (AI)

Software that performs tasks that normally need human judgment, such as understanding language, spotting patterns or making recommendations.

Automation

Using software to carry out a routine task without someone doing it by hand each time, such as sending a reminder when an invoice is overdue.

B

Bias

A pattern in an AI system's output that unfairly favors or disadvantages something, usually because of the information it learned from or how it was built.

C

Chatbot

A program people talk to in everyday language, by text or voice, to get answers or get something done.

Computer vision

AI that interprets images and video, such as reading text from a photo or spotting a defect on a production line.

Context window

How much text a language model can consider at one time, including the question, the instructions and any documents it is given.

Copilot

An AI assistant built into the tools people already use, offering suggestions and drafts while the person stays in control.

D

Data pipeline

The steps that collect information from where it is created, clean it up and deliver it to where it is used, usually automatically.

E

Embeddings

A way of turning text, images or other content into numbers that capture its meaning, so software can find things that are similar in meaning.

F

Fine-tuning

Further training of an existing AI model on a specific set of examples so it performs better on a particular kind of task.

G

Generative AI

AI that creates new content, such as text, images, audio or code, from a request.

Guardrails

Limits built around an AI system to keep it within safe and intended behavior, such as topics it will not discuss or actions it cannot take on its own.

H

Hallucination

When an AI model produces something that sounds confident but is wrong or made up. It is why important output should be checked.

Human in the loop

A design in which a person reviews or approves what an AI system produces before it has an effect, such as before a message is sent to a customer.

I

Inference

The moment an already trained AI model is used to produce an answer or prediction from new input.

Integration

Connecting two or more systems so they share information or work together, so people don't copy it between them by hand.

L

Large language model (LLM)

An AI model trained on very large amounts of text that can understand and write language. It is the technology behind most AI chat assistants.

M

Machine learning

A way of building software that learns patterns from examples instead of following only rules written by hand.

Multimodal

Describes an AI model that can work with more than one kind of input or output, such as text, images and audio.

N

Natural language processing (NLP)

The area of AI concerned with understanding and producing human language.

P

Predictive analytics

Using past information and statistics to estimate what is likely to happen, such as which customers may stop buying.

Prompt

The instructions or question given to an AI model. A clear prompt, with context and the result wanted, gives better output.

Proof of concept (pilot)

A small, limited trial that tests whether an idea works in the real business before committing to a larger rollout.

R

Retrieval-augmented generation (RAG)

A technique where an AI model first looks up relevant information from a trusted source, such as a company's documents, and then answers using it.

Robotic process automation (RPA)

Software that copies the clicks and keystrokes a person would use in an application to carry out repetitive, rule-based tasks.

S

Search engine optimization (SEO)

Improving a website so it is easier for search engines to understand and more likely to appear when people search.

Shadow AI

AI tools people use at work without the business knowing or approving, which can create privacy and accuracy risks.

Structured data (schema markup)

Labels added to a web page in a standard format so search engines and AI systems can tell what its content is, such as a business, an article or a question and its answer.

T

Token

The small pieces of text, often parts of words, that a language model reads and writes. Usage and limits are often counted in tokens.

Training data

The examples an AI model learns from. Its quality and coverage shape what the model does well and where it fails.

V

Vector database

A database designed to store embeddings and quickly find the items closest in meaning to a query.

W

Workflow orchestration

Coordinating the steps, tools and approvals of a multi-step process so they run in the right order and failures are handled.