AI terms, explained in plain language.

A step-by-step set of rules a computer follows to solve a problem or make a decision. In AI, it often means the method behind a model.

The trained system that takes an input and produces an output, such as a prediction, summary, answer, or classification.

A branch of AI where systems learn patterns from data instead of being programmed with only fixed rules.

A form of machine learning that uses many layers of computation to model complex patterns in text, images, audio, and other data.

A model made of connected layers of artificial "neurons" that transform inputs into outputs. Neural networks power much of modern AI.

A model trained on very large amounts of text so it can understand and generate language. ChatGPT-style systems are built on LLMs.

A model designed to spend more effort working through multi-step problems, planning, or analysis before answering.

A smaller language model designed to be cheaper, faster, or easier to deploy than a frontier LLM, often for narrower tasks.

AI that creates new content, such as text, images, audio, video, or code, based on patterns learned from training data.

AI that can work across more than one type of input or output, such as text plus images, or audio plus text.

The branch of AI focused on understanding and generating human language.

AI techniques for understanding images and video, such as classifying, detecting, or extracting information from visuals.

A pattern where the system first retrieves relevant information from a knowledge base, then asks the model to answer using that material.

A broader family of approaches that enrich generation with structured knowledge, often using curated facts or a knowledge graph. The label is less standardized than RAG, but the goal is the same: make answers more grounded.

A numerical representation of meaning. Embeddings let systems compare whether two pieces of text are similar, even if they use different words.

The list of numbers produced by an embedding model that captures the meaning of text, images, or other content in machine-readable form.

A database optimized for storing embeddings and finding the most semantically similar items quickly.

Search based on meaning instead of exact keyword matching. It helps an AI system find relevant content even when the wording differs.

Breaking long documents into smaller sections so they can be embedded, stored, retrieved, and cited more effectively.

Extra structured information about a file or record, such as source, date, product line, customer segment, or document type.

A second pass that reorders retrieved results so the most useful or most relevant sources are shown to the model first.

A standard way for AI applications to discover and use tools, files, and external systems. In plain terms, it helps models work with business context more safely and consistently than copying everything into a prompt.

Data without a fixed schema, such as emails, PDFs, call recordings, images, or free-form notes.

The data used to teach a model during training or fine-tuning.

Artificially generated data designed to resemble real data for training, testing, or privacy-sensitive use cases.

The process of collecting data from one system, cleaning or reshaping it, and loading it into another system for analysis or application use.

A central store for cleaned, structured business data used for reporting, analytics, and decision-making.

A storage layer that can hold large amounts of raw structured and unstructured data before it is fully organized.

A defined interface that lets one system send requests to another system and get structured responses back.

A package of tools, code, and documentation that makes it easier for developers to build on top of an API or platform.

An automatic event notification sent from one system to another when something happens, such as a lead being created or a payment succeeding.

Technology that turns text in scanned documents, screenshots, or images into machine-readable text.

Software automation that mimics repetitive computer actions like clicking, copying data, or moving information between systems.

Using software to move a process forward automatically, such as routing leads, answering standard questions, creating tickets, or syncing records.

The sequence of steps a business process follows from start to finish. In AI discussions, this usually means where the model fits inside a real operational process.

Coordinating multiple steps, systems, or tools so a larger workflow runs in the right order.

Information that can identify a real person, such as name, phone number, email address, social security number, or account details.

How understandable a model's output or decision process is to humans.

When a model learns the training data too closely and performs worse on new, unseen examples.

A testing method that repeatedly splits data into training and evaluation subsets to check whether performance is consistent.

The percentage of predictions that were correct overall.

Of the predictions the model labeled positive, the share that were actually correct.

Of the real positive cases, the share the model successfully found.

A structured test used to check whether a model or AI workflow performs well enough for a real use case.

A standardized test or score used to compare model performance across systems.

Stress-testing an AI system by trying to break it, exploit it, or expose unsafe behavior before real users do.

A technique for manipulating an AI system by slipping in malicious or conflicting instructions through user input, retrieved content, or connected tools.

AI-generated or AI-altered media, often audio or video, designed to look or sound like a real person or event.

Rules, checks, or safety layers added around an AI system to reduce harmful, unsafe, or off-policy output.

How personal or sensitive information is collected, stored, shared, and protected when AI systems are used.

Meeting the legal, contractual, or industry-specific requirements that apply to how data and AI systems are handled.

The practice of developing and using AI in ways that are fair, safe, transparent, accountable, and aligned with user rights.

The policies, controls, and accountability structures an organization uses to guide how AI is built, approved, monitored, and updated.