Digital know-how for everyoneThe everyday AI guide
AI talk can sound like another language. Here are 20 useful terms, familiar examples, and three tiny decisions to practice. No technical background needed.
Made for staff and curious neighbors. Every example is fictional.
Use approved tools. Follow your organization’s policy.
Protect people’s information. Practice with made-up examples.
Check before sharing. You still own the final review.
01 / A small dictionary for a big topic
Start anywhere. Each card has one short explanation and a pretend public-health example. These are learning examples, not recommendations to put real records into an AI service.
Matches terms, explanations, and examples. Search stays in this page.
Try a shorter word or choose another group.
The big ideas behind the buttons.
Computer systems that perform tasks such as recognizing patterns, making predictions, or producing content.
Picture this: A fictional outreach team uses a tool to sort made-up questions into topics.
Reference: NIST definitionA way to build AI by learning patterns from examples instead of writing a rule for every case.
Picture this: A practice model learns from invented attendance records to estimate event turnout.
Reference: technical glossaryAI that creates content, such as text, images, audio, or video, from patterns it has learned.
Picture this: It drafts three titles for a pretend community walking event.
Reference: NIST generative AI profileThe learned pattern system that takes input and produces a result. It is one part of an AI tool.
Picture this: Two models produce different drafts from the same pretend event notice.
Reference: generative AI glossaryThe process of adjusting a model using examples so it can learn patterns.
Picture this: A practice system learns from invented messages labeled “event” or “general question.”
Reference: technical glossaryUsing a trained model to produce a result for a new input.
Picture this: The trained practice system sorts a new made-up outreach question.
Reference: generative AI glossaryA model that estimates an outcome from patterns in data. An estimate is not a guarantee.
Picture this: A fictional team estimates chairs needed from invented event attendance totals.
Reference: technical glossaryAble to work with more than one kind of information, such as text and images.
Picture this: A tool describes a made-up poster and suggests a shorter headline.
Reference: generative AI glossaryUnderstand what goes in, what comes out, and what happens next.
The instructions or information you give an AI tool to guide its response.
Picture this: “Rewrite this fictional event notice in plain language. Keep the supplied date unchanged.”
Reference: generative AI glossaryThe result an AI system returns, such as an answer, image, draft, or prediction.
Picture this: A rewritten pretend event notice is an output, ready for a person to review.
Reference: generative AI glossaryThe amount of information a model can work with at one time. It has a limit.
Picture this: A long fictional planning chat may not keep every earlier detail available.
Reference: technical glossaryA small unit a model processes. For text, it may be a word, part of a word, or punctuation.
Picture this: A pretend flyer uses many tokens; the token count is not simply its word count.
Reference: generative AI glossaryGrounding connects answers to reference information. Retrieval-augmented generation (RAG) fetches relevant material for the model to use.
Picture this: A practice assistant looks up an approved public flyer before drafting a reply. You still check the source.
Reference: generative AI glossaryAutomation performs steps for you. An AI agent can plan steps and use tools to act toward a goal.
Picture this: A practice agent prepares a pretend reminder, but a person reviews it before anything is sent.
A drafting tool and a tool allowed to send messages have different consequences. Know what it can access or change, and use the approval steps your organization requires.
The habits that matter more than knowing every acronym.
An AI answer that invents or misstates information while presenting it as if it were true.
Picture this: A tool invents opening hours for a fictional clinic instead of saying it does not know.
Open the cited source and check that it actually supports the claim. A polished sentence or realistic-looking reference is not verification.
Patterns in data, design, or use that can produce unfair or uneven results for different people.
Picture this: A pretend campaign draft assumes every resident has a car and reliable internet access.
Check language, accessibility, assumptions, and whose experiences are missing. A neutral-sounding answer can still overlook people. Ask the appropriate reviewer to assess the effect.
Protecting information about people and controlling how it is collected, used, and shared.
Picture this: Practice with “Fictional Person A,” not a real patient’s name, visit notes, or contact details.
Follow organizational policy and use a tool approved for that specific information and purpose. Removing a name alone may not make a record safe to share. When unsure, ask your designated privacy or IT contact.
A responsible person checks whether an AI result is accurate, appropriate, and safe for its intended use.
Picture this: A reviewer checks the date, audience, and wording before a pretend outreach flyer becomes a real draft.
Reference: NIST AI Risk Management FrameworkAI-created or altered audio, images, or video that can imitate a real person or event.
Picture this: A pretend caller sounds like a supervisor. Staff verify an unusual request through a known contact.
Reference: NIST generative AI profileAn answer can sound certain and still be wrong. A stated confidence score is not proof that it is correct.
Picture this: “Definitely open Saturday” still needs checking against the fictional clinic’s official schedule.
Reference: NIST generative AI profile02 / A little practice
Three fictional situations. Choose a response, get an explanation, and try again if you like. No scores are saved or sent.
An approved AI tool drafts a notice for a pretend health fair. It added a location and a registration link you did not supply.
Check the location, link, date, and claims against the official event information. A second confident AI answer is not an independent check.
A free AI website offers to summarize visit notes. Your organization has not approved it for patient information.
Keep patient information out of an unapproved tool. Follow organizational policy and ask your designated privacy or IT contact about the right workflow.
An AI says the pretend clinic opens on Saturday. The current official schedule says it is closed.
Use the current official schedule. If the source itself is unclear, ask the responsible team. Confidence is not a substitute for evidence.
03 / Keep learning
These are plain-language summaries, not formal definitions or an approved-tool list. Follow your organization’s current policy. References checked October 26, 2026.
A starting point for the broad term “AI.”
Human oversight, context, and trustworthy use.
Confabulation, information integrity, and checking sources. See sections 2.2 and A.1.6, and action MS-2.5-003.
Understanding and managing risks to people from data use.
How harmful bias can arise across technology and its use.
Machine learning, training, context windows, predictions, and agents.
Models, prompts, tokens, inference, multimodal systems, and grounding. A technical reference, not a product recommendation.