/12

AI Glossary

Common AI terms explained in one or two plain-English sentences with no requirement to sound technical afterward.

Start here

7 min

Open +

Enough vocabulary to follow the meeting. Not enough vocabulary to become the person who says "synergistic agentic orchestration layer."

Leave knowing / Understand the words

Plain English

The words you will keep hearing.

AI

A broad label for computers doing tasks that look like perception, language, prediction, or decision-making.

Machine learning

Software that learns patterns from examples instead of having every rule written by hand.

LLM

Large language model. A model trained on huge amounts of data to work with language.

Model

The learned pattern engine that turns input into a predicted output.

Prompt

The job, question, or instruction you give the model.

Context

Everything the model can see for this request: instructions, history, files, tool results, and supplied information.

Token

A small chunk of text a model reads and writes. Usage and context limits are often measured in tokens.

Context window

How much information a model can consider at once.

Hallucination

A confident-looking answer that is unsupported, invented, or wrong.

Inference

The moment a trained model is actually run to produce an answer.

Training

The process of changing a model using data so it learns useful patterns.

Fine-tuning

Additional training that pushes a model toward a narrower behavior or style.

Embedding

A numerical representation that helps computers compare meaning or similarity.

RAG

Retrieval-augmented generation. Fetch source material first, then ask the model to answer with that context.

Tool use

Letting AI call a real capability like search, a database, a calculator, code, or an API.

Function calling

A structured way for a model to request one of those tools with specific inputs.

Agent

A system where AI can use tools and repeat steps toward a goal inside rules and permissions.

Workflow

A known sequence of steps. Not every workflow needs an agent.

Multimodal

Able to work with more than one kind of input or output, such as text, images, audio, or video.

Reasoning

The model using intermediate work or structured steps to solve a harder task. It still needs verification.

Guardrail

A rule or system boundary that restricts unsafe or invalid behavior.

Eval

A repeatable test that measures whether an AI system does the job well enough.

Benchmark

A standardized test used to compare model or system performance.

Latency

How long the system takes to respond or finish a task.

Human in the loop

A person reviews, approves, corrects, or handles cases the system should not decide alone.

MCP

Model Context Protocol. A standard way for AI applications to discover and use tools or connected context.

Jargon translator

Make the next AI sales deck speak normal human.

Try the template

I am going to paste AI language from a vendor, article, meeting, or proposal.

Translate it into plain English. For every technical or vague term:
1. Say what it means in one normal sentence
2. Say what capability it would actually give us
3. Flag anything that sounds like marketing rather than a specific capability
4. Give me one question I should ask before believing the claim

Do not make the language more technical. If a term has no single agreed meaning, say that.

Text:
[PASTE THE JARGON]

Notes / Subscribe

No schedule. No content calendar. A note goes out only when there is something worth understanding.