AI Concepts, Plainly Explained

The broader category of AI systems that act with some independence toward goals, including agents, groups of agents and the workflows around them. It is often used as an adjective, as in “agentic experience”. An agentic system does more than answer once: it plans, hands steps to others, reviews the results and goes again until the goal is met, with a person approving at the key moments.

Also called: Agentic system, agentic

Sources: Anthropic (opens in a new tab) Vendor; Gartner (opens in a new tab) Analyst

See also: AI Agent, Human in the Loop, Orchestrator, Workflow

Calling an AI assistant an “agent” when it does not act independently. Gartner calls this the most common misconception about AI agents.

Source: Gartner (opens in a new tab) Analyst

See also: AI Assistant, AI Agent

An AI system given a goal that works toward it with some independence. It decides which steps to take and which tools to use, and acts, pausing for a person at checkpoints or when it is blocked. On the AI Thoughts page each agent gets one narrow role, the fewest tools that do the job and a set of standing rules. Few agents beat many: split the work only to protect context, to run steps in parallel, to specialize or to add a checker.

Also called: Agent

Sources: Anthropic (opens in a new tab) Vendor; Gartner (opens in a new tab) Analyst; Microsoft (opens in a new tab) Vendor

See also: Agentic AI, Constitution, Least Privilege, Orchestrator, Validator, Workflow

An AI system that helps a person with tasks in response to what that person asks. It depends on human direction and does not act on its own. Gartner places assistants built into products as the first stage of enterprise AI, before task-specific agents.

Sources: Gartner (opens in a new tab) Analyst; AWS (opens in a new tab) Vendor

See also: Copilot, AI Agent, Agentwashing

The components, patterns, tokens and guidance a company uses for AI features specifically: AI labels, loading and uncertainty states, feedback controls, hand-off patterns and disclosure patterns.

Also called: AI pattern library

See also: AI Label, Design Tokens, Escalation

A message that informs the user and asks nothing of them. This is my working definition, the counterpart to a disclosure.

See also: AI Disclosure

A message that tells the user something and asks them to act on it: acknowledge it, consent or choose. That is what sets it apart from a disclaimer. This is my working definition; there is no single standard one.

See also: AI Disclaimer, Consent, EU AI Act, Article 50

A visual marker that identifies AI-generated content or an AI feature. IBM’s Carbon for AI guidance says the AI label should never be disabled.

Source: IBM Carbon for AI (opens in a new tab) Vendor

See also: AI Disclosure, EU AI Act, Article 50

My working definition: a design process built around AI tools from the start, generating, testing and governing with them, rather than one where AI is bolted onto the old steps. “AI-native” is industry usage, not a defined standard.

Source: Industry usage Practitioner

See also: AI Design System, Lift and Shift

A point where an agent must stop and get a person’s yes before acting. Gates are often tiered: act automatically, act and notify, or wait until approved.

Sources: Salesforce Agentforce Vendor; Microsoft HAX Toolkit (opens in a new tab) Vendor

See also: Human in the Loop, Levels of Autonomy

The record of every call the system makes. It is added to and never edited, so legal and security teams can see what happened. The gateway writes to it, and the pipeline adds its own events, such as the approval and the final ship, so one trail covers both.

Also called: Audit trail

See also: Gateway, Pipeline (Continuous Integration)

A large language model given extra abilities: retrieval (looking things up), tools (taking actions) and memory. It is the basic building block of every agentic system.

Source: Anthropic (opens in a new tab) Vendor

See also: Memory, RAG (Retrieval-Augmented Generation), Tool Use

A California law that requires large generative AI providers to offer visible disclosures and embedded (hidden) markings in the content their systems create. AB 853 moved its start date, and it has applied since 2 August 2026.

Also called: SB 942, amended by AB 853

Source: California Legislature, SB 942 and AB 853 Standard

See also: AI Label, EU AI Act, Article 50

A workspace beside the chat where longer work, such as a document, code or a design, lives and can be edited. OpenAI and Google call it Canvas and Anthropic calls it Artifacts. There is no single industry term.

Also called: Artifacts (Anthropic’s name)

Source: OpenAI, Google and Anthropic Vendor

See also: Generative UI

Software that holds a text conversation. Historically it was rule-based: scripted flows, decision trees, button menus and keyword matching. Modern chatbots built on large language models can follow the context of a conversation.

Watch out: The word now covers both the scripted kind and products like ChatGPT, so it carries the scripted meaning whether the writer wants it or not.

Source: AWS (opens in a new tab) Vendor

See also: Conversational AI, AI Assistant

Colorado’s first AI law (SB 24-205) was repealed and replaced by SB 26-189, signed on 14 May 2026 and taking effect on 1 January 2027. The new law takes a narrower approach built on notices to consumers.

Also called: SB 24-205, replaced by SB 26-189

Source: Colorado General Assembly, SB 26-189 (opens in a new tab) Standard

See also: AI Disclosure, Responsible AI

The user’s explicit agreement before something happens, such as their data being used, the screen changing or an action being taken on their behalf.

See also: AI Disclosure, Approval Gate

My term for the standing rules loaded into an agent before any request: what it is, what it may do, what it must never do and when to stop and ask.

See also: AI Agent, Guardrail

The umbrella category for any technology people interact with through natural conversation, in text or voice. It includes chatbots, virtual agents, voice assistants, IVR, AI assistants and copilots.

Watch out: Accurate as a category, vague as a description. Saying a product “is conversational AI” tells the reader it talks, not what it does, so many readers will picture a support chatbot or a phone menu.

Sources: IBM (opens in a new tab) Vendor; AWS (opens in a new tab) Vendor

See also: Chatbot, AI Assistant, Copilot

An AI assistant built into the application where the work happens, usually in a side panel, that can see the current context (the page, document or record) and help in place.

Watch out: “Copilot” is also Microsoft’s product brand, and GitHub’s. Lowercase “copilot” is the general pattern.

Source: Microsoft (opens in a new tab) Vendor

See also: AI Assistant

The named values of a design system, such as color, type and spacing, kept as data. On the AI Thoughts page they live in a design-tokens.json file in Git, in the Design Tokens Community Group format, and sync into Figma.

See also: AI Design System, Drift, Prop Contract

A check run by code that gives the same answer every time, so a model cannot argue its way past it. It is used for the rules that must never fail: required AI disclosures, accessibility, brand and performance budgets.

See also: Guardrail, Non-Deterministic, Pipeline (Continuous Integration), Validator

When two places that should agree, such as Figma and the code, stop matching. A parity check flags the difference, and a person decides which side wins.

See also: Design Tokens, System of Record

A way of turning text into meaning-coordinates, so passages with similar meaning sit close together and can be found by search.

See also: RAG (Retrieval-Augmented Generation), Vector Database

Moving a conversation from the AI to a person. A warm transfer passes the context along, so the user doesn’t have to repeat themselves. A cold transfer starts over.

Also called: Hand-off, warm or cold transfer

Source: Contact-center industry usage Practitioner

See also: Human in the Loop, Virtual Agent

The part of the EU AI Act that requires people to be told when they are interacting with an AI system, and requires AI-generated content to be marked. It applies from 2 August 2026. The 2026 Digital Omnibus gave one extension: generative AI systems already on the market before that date have until 2 December 2026 for the marking and detection duty only.

Also called: Transparency obligations

Sources: Regulation (EU) 2024/1689 (opens in a new tab) Standard; European Commission (opens in a new tab) Standard

See also: AI Disclosure, AI Label, California AI Transparency Act

A setup where each kind of information has one clear owner, instead of one single source of truth. When two places disagree, the system points it out and a person decides.

See also: Drift, Skin and Skeleton, System of Record

The one doorway every AI request passes through, run by the team that owns it. It sets permissions and scopes, watches every request and keeps the audit log. The MCP standard leaves those jobs to whoever builds with it, so the gateway is a layer we add.

Also called: MCP gateway

See also: Audit Log, Least Privilege, MCP (Model Context Protocol)

AI models that generate new content (text, images, audio, video or code) modeled on the data they learned from. NIST defines it as models that copy the structure of their input data to produce new, synthetic content.

Also called: GenAI, GAI

Source: NIST AI 600-1 (opens in a new tab) Standard

See also: Large Language Model (LLM)

Interface the model produces or assembles while it runs, rather than text alone.

Source: Vercel AI SDK Vendor

See also: In-Chat Tool, Canvas

Tying a model’s answer to a specific source the user can check, such as a document, a record or the current page. Retrieval-augmented generation is one way to ground an answer.

Source: Industry usage, no single governing definition Practitioner

See also: Hallucination, RAG (Retrieval-Augmented Generation)

A rule that must hold every time. Guardrails act at three points: the constitution inside each agent, the gateway’s access policy and the pipeline’s checks on the output. Knowledge informs; guardrails enforce.

See also: Constitution, Deterministic Check, Gateway, Pipeline (Continuous Integration)

Content a model states with confidence that is false. NIST warns that it can mislead or deceive users.

Also called: Confabulation (NIST’s preferred term), fabrication

Source: NIST AI 600-1 (opens in a new tab) Standard

See also: Grounding, RAG (Retrieval-Augmented Generation)

A person takes part in the decision while the system runs, reviewing, approving, correcting or rejecting before an action takes effect. By design, the person sets the goal in plain language, approves at defined checkpoints (not only at the end) and decides what “good” means. No standards body has one definition of it for product design.

Also called: HITL

Sources: Credo AI (opens in a new tab) Practitioner; Anthropic (opens in a new tab) Vendor

See also: Approval Gate, Human on the Loop, Intent, Levels of Autonomy

The system acts on its own, and a person supervises: they watch the outcomes and can step in or correct, without approving each action.

Also called: HOTL

Source: Credo AI (opens in a new tab) Practitioner

See also: Human in the Loop, Levels of Autonomy

An interactive piece of interface, such as a form, a ticket or a picker, that appears inside the conversation instead of sending the user somewhere else. They are often micro-frontend tools backed by MCP servers. The name varies by vendor.

Also called: Embedded UI

Sources: Anthropic Vendor; OpenAI Apps SDK Vendor

See also: Generative UI, MCP (Model Context Protocol)

A goal stated in plain language, with no ticket and no formal spec. The plan, the research and the building are all worked out from it.

See also: Human in the Loop, Intent Detection, Orchestrator

Working out what the user is asking for and sending the request to the right tool or data source. Teams often tune it with trigger phrases, the wordings that should send a request to each tool.

Also called: Routing, trigger phrases

See also: Intent, Tool Use

A phone system that lets callers use their voice or keypad to move through menus, do self-service tasks or reach a person. Traditional IVR uses keypad tones (“press 1”). Conversational IVR uses natural language understanding, so callers can just say what they need.

Also called: Conversational IVR

Source: Conversation Design Institute (opens in a new tab) Practitioner

See also: Conversational AI, Escalation

A generative model trained on very large amounts of text that predicts and produces language. Claude, GPT-4o and Gemini are large language models. On the AI Thoughts page one model plays every role: the orchestrator, the agents and the validator.

See also: AI Agent, Generative AI, Non-Deterministic

Giving an agent the fewest tools and the narrowest access that do the job.

See also: AI Agent, Gateway

A scale from “the person does everything” to “the system acts alone”, used to decide which human-control pattern a feature needs. Several competing scales exist. One common version has five steps, L0 to L4.

Source: Several competing scales Practitioner

See also: Approval Gate, Human in the Loop, Human on the Loop

My working philosophy: borrow patterns people already trust, then honor what those patterns promise. In IT, “lift and shift” also means moving an application to the cloud unchanged; the context makes the meaning clear.

See also: AI-Native Design Process

An open standard for connecting AI applications to outside tools, data and workflows, often described as a universal port. It is commonly said to turn M×N custom integrations into M+N. Anthropic created it, and Claude, ChatGPT, VS Code, Cursor and others now support it.

Source: modelcontextprotocol.io (opens in a new tab) Vendor

See also: Gateway, In-Chat Tool, Tool Use

What an AI system keeps between turns of a conversation (conversation memory) or between sessions (long-term memory), and the controls that let the user see and change it. On the AI Thoughts page, memory keeps past decisions and outcomes for the next request, so it starts better informed.

Sources: Google PAIR (opens in a new tab) Vendor; Microsoft Vendor

See also: Augmented LLM, RAG (Retrieval-Augmented Generation)

The same input can produce a different output on a different run. Large language models work this way by default.

Why it matters for design: Designers can’t design for one fixed answer. They design for a range of answers, for uncertainty and for what the user does when an answer is wrong.

See also: Deterministic Check, Hallucination

The agent that turns a goal into a plan, hands each step to other agents, gathers the results and goes again until the goal is met. It pauses at the person’s checkpoints and writes every action to the log.

See also: AI Agent, Audit Log, Intent

The automated checks that run on the output before anything ships: disclosures, accessibility, brand rules and performance budgets. It also writes its own events, such as the approval and the final ship, to the audit log.

Also called: CI

See also: Audit Log, Deterministic Check, Guardrail

The system notices something and brings it to the user before being asked, rather than waiting for a question.

Also called: Proactive insights, nudges

Sources: Google PAIR (opens in a new tab) Vendor; Shape of AI (opens in a new tab) Practitioner

See also: AI Agent, Human on the Loop

A machine-readable description of what a component allows, generated from the code. It is why no separate mapping layer is needed between design and code.

See also: Design Tokens

Before answering, the system retrieves relevant documents or records and gives them to the model, so the answer rests on that material rather than only on what the model memorized in training. Documents are split into chunks, turned into embeddings and stored in a vector database, and each request retrieves only the most relevant passages. It is an open-book exam instead of a closed one.

Source: Lewis et al., NeurIPS 2020 (opens in a new tab) Standard

See also: Embedding, Grounding, Memory, Vector Database

The practice of designing, building and governing AI so it is safe, fair, transparent and accountable. It is a field, not a single standard.

Sources: NIST AI Risk Management Framework (opens in a new tab) Standard; Microsoft and Google frameworks Vendor

See also: AI Disclosure, AI Label

Two kinds of design-system pattern. The Skeleton is the working components, owned by engineering. The Skin is the brand and identity layer (color, type, voice and rules), owned by design. One Skin can dress many Skeletons.

See also: Design Tokens, Federated Model

The one place that is the authority for a particular kind of information. In a federated model each kind of information has its own.

See also: Drift, Federated Model

The model asks the application to run a function, such as searching a database, opening a ticket or checking a status, and then uses the result. The model decides when to call a tool from the user’s request and the tool’s description.

Also called: Function calling

Source: Anthropic (opens in a new tab) Vendor

See also: Augmented LLM, MCP (Model Context Protocol)

A Utah law, in effect since 1 May 2024, that requires disclosure when consumers interact with generative AI. Amendments in 2025 narrowed when that disclosure is required.

Also called: SB 149

Source: Utah Legislature, SB 149 (2024) (opens in a new tab) Standard

See also: AI Disclosure

A separate agent whose only job is to check the work against the rules, using deterministic checks. Because it is kept apart from the agent that made the work, the maker cannot talk it out of a failure.

Also called: Checker

See also: AI Agent, Deterministic Check

Where embeddings are stored, so a request can find the passages closest in meaning to what it needs.

See also: Embedding, RAG (Retrieval-Augmented Generation)

Contact-center and customer-service terms for an automated conversational system that handles customer requests, often with a hand-off to a person. Gartner used “virtual customer assistant” for this class.

Watch out: Two collisions. Here “agent” means a customer-service seat, not an autonomous AI agent. And in the same contact center, “agent” also means the human support person.

Also called: Virtual assistant, virtual customer assistant (VCA)

Sources: IBM (opens in a new tab) Vendor; Gartner (opens in a new tab) Analyst

See also: AI Agent, Escalation

A system where a large language model and its tools follow a path someone designed in advance. An agent chooses its own path; a workflow follows one.

Why it matters for design: Most “agents” in enterprise products are partly workflows. Knowing which parts are fixed tells the designer where the user needs to see a decision and where they need to see progress.

Source: Anthropic (opens in a new tab) Vendor

See also: AI Agent, Agentic AI