AI Terms Glossary
A working reference for the language of AI in entry-level work. 30 essential terms mapped to the 7-pillar framework, with maturity levels for what to learn now vs. later.
Acceptable Use Policy (AUP)
An organization's documented rules for how employees may use AI tools — specifying what data can be entered, which tools are approved, and what outputs require review or disclosure.
Why this matters
Violating an employer's AUP with AI tools can constitute a policy breach. Students should ask about AUPs before using AI at work.
In the framework
This term spans Pillar 4 · Ethics and Pillar 6 · Data & Privacy.
Adversarial Testing
Deliberately crafting inputs designed to expose weaknesses, errors, or inconsistencies in AI outputs — probing edge cases, unusual phrasings, or misleading contexts to assess reliability before building workflows around a tool.
Why this matters
Adversarial testing is how professionals validate AI tools before committing to them. It's critical thinking applied as a structured method.
In the framework
This term lives in Pillar 3 · Critical Thinking. Explore the full pillar or take the assessment to see where you stand on it.
Agentic AI
AI systems that can plan and execute multi-step tasks autonomously — browsing the web, running code, sending emails, or interacting with external services — without step-by-step human instruction for each action.
Why this matters
Agentic AI is the next major shift in how AI is used in business. Understanding it now is a competitive advantage.
In the framework
This term spans Pillar 2 · Interaction and Pillar 7 · Future Readiness.
Agentic Workflow Design
The discipline of designing multi-step AI workflows in which agents take autonomous actions — specifying task sequences, guardrails, handoff points, and human review triggers to ensure reliable, auditable outcomes.
Why this matters
As agentic AI moves into business operations, the ability to design and govern these workflows is becoming a high-value professional skill.
In the framework
This term spans Pillar 2 · Interaction and Pillar 5 · Application.
AI Agent Framework
A software platform or library — such as LangChain, AutoGen, or CrewAI — that provides tools for building, orchestrating, and deploying multi-agent AI systems, abstracting the complexity of connecting models, tools, and memory.
Why this matters
Agent frameworks are moving from developer tools to business infrastructure. Understanding what they are provides context for evaluating AI platform investments and vendor claims.
In the framework
This term spans Pillar 2 · Interaction and Pillar 7 · Future Readiness.
AI Audit
A systematic review of an AI system's inputs, outputs, decision logic, and outcomes — conducted to assess accuracy, bias, compliance, and alignment with stated policies or ethical standards.
Why this matters
AI audits are becoming a compliance requirement in regulated industries. Entry-level employees in finance, HR, and healthcare will encounter them.
In the framework
This term lives in Pillar 4 · Ethics. Explore the full pillar or take the assessment to see where you stand on it.
AI Co-pilot
An AI tool or feature integrated into a professional application — like Microsoft Copilot in Word or Notion AI — designed to assist with tasks within an existing workflow rather than as a standalone chat interface.
Why this matters
Co-pilots are how most students will encounter AI in their first jobs. Knowing how to use them is a real workplace skill.
In the framework
This term spans Pillar 2 · Interaction and Pillar 5 · Application.
AI Disclosure
The practice of informing relevant parties — employers, professors, clients, or audiences — that AI tools were used to assist in creating work. Disclosure norms vary by context and organization.
Why this matters
Academic integrity and professional ethics both increasingly require knowing when and how to disclose AI use.
In the framework
This term lives in Pillar 4 · Ethics. Explore the full pillar or take the assessment to see where you stand on it.
AI Ethics
The study and application of moral principles to AI development and use. Covers fairness, accountability, transparency, and the potential for harm — both intended and unintended.
Why this matters
Every professional who uses AI tools has an ethical responsibility, regardless of their role or title.
In the framework
This term lives in Pillar 4 · Ethics. Explore the full pillar or take the assessment to see where you stand on it.
AI Governance
The policies, processes, and oversight structures that organizations use to manage how AI is developed, deployed, and used — covering ethics, compliance, risk, and accountability.
Why this matters
Graduates entering large organizations will be expected to understand and operate within AI governance frameworks.
In the framework
This term lives in Pillar 4 · Ethics. Explore the full pillar or take the assessment to see where you stand on it.
AI Overreliance
The tendency to accept AI outputs uncritically — reducing independent judgment, verification, and creative thinking because the AI response feels sufficient. Overreliance degrades skill development and increases error risk.
Why this matters
Research shows AI overreliance is a documented professional risk. Knowing the concept helps you build intentional habits around maintaining your own judgment.
In the framework
This term spans Pillar 3 · Critical Thinking and Pillar 4 · Ethics.
AI Red Teaming
A structured discipline in which adversarial teams systematically probe AI systems for vulnerabilities, harmful behaviors, and failure modes — combining technical testing, social engineering simulations, and policy violation attempts to produce a comprehensive risk profile.
Why this matters
AI red teaming is becoming a professional specialization. Understanding it provides context for how organizations manage AI risk before and after deployment.
In the framework
This term spans Pillar 3 · Critical Thinking and Pillar 4 · Ethics.
AI Safety
A research area and design discipline focused on ensuring AI systems behave reliably, interpretably, and in alignment with human values — especially as capabilities increase. Distinct from cybersecurity, though related.
Why this matters
AI safety is shaping how models are developed and regulated. Business and communications students benefit from knowing the discourse.
In the framework
This term spans Pillar 1 · Foundations and Pillar 4 · Ethics.
AI Watermarking
Techniques for embedding imperceptible signals into AI-generated content — text, images, or audio — that allow it to be identified as AI-generated even after editing or distribution.
Why this matters
AI watermarking is increasingly relevant for content authentication, copyright, and disclosure compliance in marketing and media roles.
In the framework
This term lives in Pillar 4 · Ethics. Explore the full pillar or take the assessment to see where you stand on it.
AI Workflow
A repeatable sequence of AI-assisted steps designed to complete a specific business or creative task more efficiently — such as using AI to research, outline, draft, and edit a report in stages.
Why this matters
Building systematic AI workflows is what separates ad-hoc AI use from genuine professional productivity improvement.
In the framework
This term spans Pillar 2 · Interaction and Pillar 5 · Application.
Algorithm
A set of rules or instructions a computer follows to complete a task or solve a problem. In AI, algorithms define how a model processes inputs and produces outputs.
Why this matters
Demystifies the 'black box' — AI follows processes, even if those processes are complex.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Algorithmic Accountability
The principle that organizations and individuals using AI systems bear responsibility for the outcomes those systems produce — including errors, biases, and harms — regardless of whether the AI operated as designed.
Why this matters
A foundational concept for anyone in a professional role where AI-assisted decisions affect other people.
In the framework
This term lives in Pillar 4 · Ethics. Explore the full pillar or take the assessment to see where you stand on it.
Alignment (AI Safety)
The challenge of ensuring that AI systems behave in ways consistent with human values, intentions, and goals — particularly as models become more capable and autonomous. A central concern of AI safety research.
Why this matters
Alignment is the foundational concept behind much of the policy and safety discourse shaping how AI is developed and regulated globally.
In the framework
This term spans Pillar 1 · Foundations and Pillar 4 · Ethics.
Anonymization (AI context)
The process of removing or obscuring personally identifiable information from data before it is used with AI tools — replacing names with codes, aggregating records, or removing identifying attributes.
Why this matters
Anonymization is the most practical data privacy technique for professionals using AI tools on real-world business data.
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
API (AI context)
Application Programming Interface — a technical interface that allows software to communicate with an AI model programmatically. APIs enable developers and businesses to embed AI capabilities into their own products and workflows.
Why this matters
Business students don't need to code, but understanding what an API is explains how AI gets integrated into the tools they'll use.
In the framework
This term spans Pillar 5 · Application and Pillar 7 · Future Readiness.
Artificial Intelligence (AI)
Computer systems designed to perform tasks that typically require human intelligence — such as understanding language, recognizing patterns, and making decisions. Modern AI is probabilistic, not deterministic.
Why this matters
The baseline concept every professional needs to understand before using AI tools in any workflow.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Attention Mechanism
The core component of transformer models that allows the model to weigh the relevance of each word or token in an input relative to every other token, enabling the model to capture long-range language dependencies.
Why this matters
Explains why LLMs can understand nuanced context — and why they still fail at certain types of logical reasoning.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Automation
Using AI or software to execute tasks that previously required human effort — from scheduling and drafting to research and data analysis. AI-powered automation is reshaping which tasks humans focus on.
Why this matters
Students entering the workforce need to understand which tasks AI can handle and where human judgment remains essential.
In the framework
This term spans Pillar 5 · Application and Pillar 7 · Future Readiness.
Automation Bias
The tendency to favor or over-trust automated or AI-generated suggestions, even when they conflict with other evidence or human judgment. A well-documented cognitive bias in human-computer interaction.
Why this matters
Automation bias is one of the most significant professional risks of AI adoption. Knowing it exists helps you counteract it.
In the framework
This term spans Pillar 3 · Critical Thinking and Pillar 4 · Ethics.
Autonomous AI Agent
An AI system capable of perceiving its environment, making decisions, and taking actions to achieve a goal with limited or no human instruction for each step. Distinguished from a chatbot by its ability to act, not just converse.
Why this matters
Autonomous agents are transitioning from research to commercial deployment. Understanding what they can and cannot do safely is critical for professional AI judgment.
In the framework
This term spans Pillar 2 · Interaction and Pillar 7 · Future Readiness.
Benchmark (AI)
A standardized test used to evaluate and compare AI model performance on specific tasks — such as reasoning, coding, or language understanding. Benchmarks are how developers and researchers measure model capability claims.
Why this matters
AI companies cite benchmark scores to market their models. Knowing what benchmarks measure — and don't measure — prevents you from being misled by capability claims.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Bias (AI)
Systematic errors in AI outputs that reflect imbalances or prejudices in training data or model design. Bias can disadvantage certain groups or produce skewed results in ways that aren't immediately obvious.
Why this matters
Relevant to both ethical use and quality control — biased outputs can cause professional and reputational harm.
In the framework
This term spans Pillar 3 · Critical Thinking and Pillar 4 · Ethics.
Brand Voice (AI context)
The practice of configuring AI tools — through system prompts, style guides, or fine-tuning — to produce content that consistently matches an organization's tone, vocabulary, and personality standards.
Why this matters
Maintaining brand voice consistency at scale is a real challenge when AI is used across a content team. Students who understand how to configure this are immediately more valuable.
In the framework
This term lives in Pillar 5 · Application. Explore the full pillar or take the assessment to see where you stand on it.
Capability Illusion
The mistaken belief that an AI tool can perform a task reliably based on isolated impressive examples — without accounting for consistency, failure rate, and edge-case behavior across real-world use.
Why this matters
Capability illusions lead professionals to overcommit AI tools to critical tasks before adequate testing. It's a specific form of AI overreliance.
In the framework
This term spans Pillar 3 · Critical Thinking and Pillar 7 · Future Readiness.
Catastrophic Forgetting
The tendency of neural networks to lose previously learned information when trained on new data — a fundamental challenge in developing AI systems that can update their knowledge without erasing what they already know.
Why this matters
Explains why continual learning in AI is technically difficult, and why models can't simply be 'updated' with new facts without significant retraining.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
CCPA
The California Consumer Privacy Act — a state privacy law granting California residents rights over their personal data, including the right to know what is collected, opt out of sale, and request deletion. Relevant to AI tools handling customer data.
Why this matters
CCPA applies to any organization with California customers — which is most businesses. Understanding its implications for AI data handling is a practical compliance skill.
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
Chain-of-Thought Prompting
A technique where you ask the AI to reason through a problem step by step before providing an answer. Improves accuracy on complex reasoning tasks by making the model's logic visible.
Why this matters
Useful for analytical tasks — and it makes AI reasoning auditable, which improves verification and trust.
In the framework
This term spans Pillar 2 · Interaction and Pillar 3 · Critical Thinking.
Chatbot
A software application designed to simulate conversation with users. AI-powered chatbots use LLMs to understand and generate natural language rather than following scripted responses.
Why this matters
The most common interface students encounter — understanding what's underneath the chat window improves how they use it.
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Confidential Data
Business information that carries restricted access or disclosure obligations — trade secrets, unreleased financial data, client contracts, personnel records. Entering confidential data into public AI tools is a material security risk.
Why this matters
Students need to know what counts as confidential before using AI tools on work tasks — the default should be caution.
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
Confabulation
The technical term for 'hallucination' — the AI confidently produces plausible-sounding but inaccurate content. The model is not lying; it is completing a pattern without grounding in verified facts.
Why this matters
Using the precise term helps distinguish AI error from deliberate deception — an important distinction in professional and ethical discussions.
In the framework
This term spans Pillar 1 · Foundations and Pillar 3 · Critical Thinking.
Constitutional AI
A training methodology developed by Anthropic in which AI models are trained to evaluate and revise their own outputs according to stated principles — rather than relying solely on human feedback at each step.
Why this matters
Constitutional AI is the methodology behind Claude's design. Understanding it provides context for how AI values and behaviors are engineered, not just emergent.
In the framework
This term spans Pillar 1 · Foundations and Pillar 4 · Ethics.
Content Generation
Using AI to create written, visual, or multimedia content — blog posts, social copy, email drafts, ad concepts, and more. AI generates drafts; human judgment refines and approves them.
Why this matters
Core to nearly every marketing and communications role. Students need to do this well and know when to edit.
In the framework
This term spans Pillar 2 · Interaction and Pillar 5 · Application.
Content Repurposing (AI)
Using AI to adapt existing content for different formats, channels, or audiences — turning a blog post into social posts, a webinar into a summary, or a report into a slide deck — without recreating from scratch.
Why this matters
Content repurposing is one of the highest-ROI AI use cases in marketing. It's a practical skill that directly reduces production time.
In the framework
This term lives in Pillar 5 · Application. Explore the full pillar or take the assessment to see where you stand on it.
Context Window
The total amount of text (measured in tokens) that an AI model can process at one time — including the prompt, conversation history, and any documents provided. Information outside the window is not accessible to the model.
Why this matters
Context window limits directly affect how much material you can send to an AI at once — a practical constraint for research and summarization tasks.
Related terms
- Token
- Long-context model
In the framework
This term spans Pillar 1 · Foundations and Pillar 2 · Interaction.
Copyright (AI)
Legal frameworks governing ownership of AI-generated content are still evolving. Key questions include who owns AI outputs, whether training on copyrighted material is permissible, and what constitutes original authorship.
Why this matters
Business and marketing students need to understand the IP implications of using AI to create content for professional or commercial use.
In the framework
This term lives in Pillar 4 · Ethics. Explore the full pillar or take the assessment to see where you stand on it.
Data Localization
Legal or regulatory requirements that data collected within a country or jurisdiction must be stored and processed within that same jurisdiction — affecting which cloud AI tools organizations can legally use.
Why this matters
Data localization requirements directly constrain which AI vendors and infrastructure options are legally available in global operations.
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
Data Minimization
The privacy principle of collecting and sharing only the minimum data necessary for a specific purpose. Applied to AI use, it means not entering more personal or confidential information into a model than the task requires.
Why this matters
A practical habit that reduces privacy risk and aligns with regulatory standards like GDPR.
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
Data Poisoning
A type of adversarial attack in which malicious or misleading data is introduced into an AI model's training set, deliberately distorting the model's behavior or outputs in harmful ways.
Why this matters
Data poisoning is a real security threat for organizations building or fine-tuning AI models. Understanding it informs smarter decisions about AI supply chain risk.
In the framework
This term spans Pillar 1 · Foundations and Pillar 6 · Data & Privacy.
Data Privacy
The right of individuals to control how their personal information is collected, used, and shared. AI tools that process or store user input create privacy obligations for anyone using them in a professional setting.
Why this matters
Understanding data privacy prevents accidental violations that could expose a company to regulatory and reputational risk.
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
Data Retention (AI)
Policies governing how long AI tools store user inputs, conversation histories, and generated outputs. Retention periods vary significantly between tools and have compliance implications for professional use.
Why this matters
Understanding retention policies determines whether AI tool use is appropriate for confidential or regulated business tasks.
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
Data Supply Chain (AI)
The full lifecycle of data used to train and operate AI systems — from collection and labeling through preprocessing, training, storage, and deployment. Each stage introduces potential privacy, quality, and bias risks.
Why this matters
Understanding where data comes from — and how it's handled — is foundational for responsible AI use in any business context.
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
Deepfake
AI-generated audio, video, or images that realistically depict real people saying or doing things they never did. Created using generative AI techniques, deepfakes are increasingly difficult to detect without specialized tools.
Why this matters
Deepfakes are a direct threat to trust in digital communications — a core competency concern for marketing and communications students.
In the framework
This term lives in Pillar 4 · Ethics. Explore the full pillar or take the assessment to see where you stand on it.
Differential Privacy
A mathematical framework for adding calibrated statistical noise to datasets so that individual records cannot be re-identified — enabling AI training on sensitive data while protecting personal information.
Why this matters
An emerging standard in enterprise AI. Knowing the concept enables you to engage credibly in data governance conversations.
Related terms
- Privacy
- Anonymization
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
Digital Twin
A virtual simulation of a real-world object, process, or system — increasingly built and operated using AI. In marketing and business contexts, customer digital twins model individual behavior for predictive personalization.
Why this matters
Digital twins represent an emerging application of AI in product development, operations, and customer experience design — relevant to future marketing and strategy roles.
In the framework
This term spans Pillar 5 · Application and Pillar 7 · Future Readiness.
Embeddings
Numerical representations of text, images, or other data in high-dimensional space, where items with similar meaning are positioned closer together. Embeddings enable AI models to measure semantic similarity between concepts.
Why this matters
Embeddings are the mechanism that allows AI to 'understand' meaning — foundational to search, recommendation, and classification tasks.
In the framework
This term spans Pillar 1 · Foundations and Pillar 7 · Future Readiness.
Emergent Capabilities
Unexpected abilities that appear in AI models at sufficient scale — tasks the model was not explicitly trained to perform but develops capacity for as it grows larger. Reasoning, translation, and code generation all emerged this way.
Why this matters
Emergent capabilities make AI development difficult to predict and are central to debates about AI's future trajectory.
In the framework
This term spans Pillar 1 · Foundations and Pillar 7 · Future Readiness.
Epistemic Calibration
The ability to accurately assess how confident you should be in a claim — your own or an AI's. Epistemic calibration means knowing when to trust, when to verify, and when to reject AI outputs based on evidence quality.
Why this matters
A core critical thinking skill that becomes essential when working with AI tools that express false confidence fluently.
In the framework
This term lives in Pillar 3 · Critical Thinking. Explore the full pillar or take the assessment to see where you stand on it.
EU AI Act
The European Union's landmark AI regulatory framework, classifying AI systems by risk level and imposing requirements for transparency, documentation, and human oversight. The first comprehensive AI law of its kind globally.
Why this matters
The EU AI Act is shaping global AI policy. Students entering marketing, finance, or HR roles at global companies will encounter its requirements.
In the framework
This term spans Pillar 4 · Ethics and Pillar 6 · Data & Privacy.
Few-shot Prompting
Including two to five examples of the desired input-output format directly in your prompt to guide the AI's response. Significantly improves consistency and quality for structured or templated outputs.
Why this matters
A practical technique for marketing and content tasks where format consistency matters — email templates, social copy, reports.
Related terms
- Zero-shot
- Prompt engineering
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Fine-tuning
The process of further training a pre-built foundation model on a smaller, domain-specific dataset to improve its performance on particular tasks or in particular contexts.
Why this matters
Fine-tuned models power many enterprise AI tools. Knowing this explains why a company's internal AI might behave differently from a general-purpose chatbot.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Foundation Model
A large AI model trained on broad, general-purpose data that can be adapted or fine-tuned for specific tasks. GPT-4, Claude, and Llama are all foundation models. Also called a base model.
Why this matters
Foundation models are the infrastructure layer beneath most AI tools students will use at work.
Related terms
- Pre-training
- Fine-tuning
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
GDPR
The General Data Protection Regulation — an EU law setting standards for how personal data must be collected, processed, and protected. It has extraterritorial reach and affects any organization handling EU residents' data.
Why this matters
Relevant for business and marketing students at any organization operating globally or handling customer data.
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
Generative AI
AI systems that create new content — text, images, audio, video, or code — rather than simply classifying or analyzing existing content. LLMs are one type of generative AI.
Why this matters
Generative AI is the category reshaping entry-level marketing, content, and communications roles.
In the framework
This term spans Pillar 1 · Foundations and Pillar 5 · Application.
Ground Truth
The verified, factually correct answer or data point against which an AI output is evaluated. In professional use, ground truth is the human-verified standard that AI outputs must be checked against.
Why this matters
Without a concept of ground truth, there's no basis for evaluating AI accuracy. It's the reference point that makes verification meaningful.
In the framework
This term lives in Pillar 3 · Critical Thinking. Explore the full pillar or take the assessment to see where you stand on it.
Grounding
The process of connecting AI outputs to verified, real-world facts, documents, or data sources to reduce hallucination and improve accuracy. Grounded AI responses are traceable to specific sources.
Why this matters
Grounding is the core mechanism behind more reliable AI tools — understanding it helps you evaluate which AI systems to trust for factual tasks.
Related terms
- RAG
- Citation
In the framework
This term spans Pillar 1 · Foundations and Pillar 3 · Critical Thinking.
Hallucination
When an AI model generates content that sounds confident and fluent but is factually incorrect or entirely fabricated. This is a structural limitation of LLMs, not a bug that gets 'fixed.'
Why this matters
One of the most critical concepts in AI literacy — knowing this prevents costly errors in professional work.
Related terms
- Grounding
- Verification
In the framework
This term spans Pillar 1 · Foundations and Pillar 3 · Critical Thinking.
Human-in-the-Loop
A design principle or workflow practice in which a human reviews, approves, or corrects AI outputs before they are acted upon. Essential for high-stakes, regulated, or public-facing tasks.
Why this matters
The default professional standard for AI-assisted work — AI drafts, humans decide.
Related terms
- Workflow design
- Review
In the framework
This term spans Pillar 3 · Critical Thinking and Pillar 4 · Ethics.
Inference (AI)
The process of running a trained AI model on new inputs to generate outputs. When you send a prompt and get a response, the model is performing inference. Distinct from training, which is how the model was originally built.
Why this matters
Clarifies the real-time nature of AI responses — and why the same model can give different answers to the same question.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Inference Attack
An attempt to extract private or sensitive information about individuals by analyzing AI model outputs — even when the model was not directly given that data as input.
Why this matters
Inference attacks demonstrate that simply avoiding direct input of sensitive data is not always sufficient to protect privacy in AI systems.
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
Inference-Time Compute
The computational resources used during output generation — distinct from training compute. Reasoning models use more inference-time compute to 'think longer' before responding, trading latency for accuracy.
Why this matters
Inference-time compute scaling is the current frontier of AI capability growth — understanding it explains why reasoning models produce more accurate results and why they cost more to run.
In the framework
This term spans Pillar 1 · Foundations and Pillar 7 · Future Readiness.
Instruction Following
The degree to which an AI model reliably does what its prompt asks — following format constraints, respecting limits, and executing multi-step instructions accurately. A key quality dimension that varies significantly across models.
Why this matters
Not all models follow instructions equally well. Understanding instruction following as a model property helps professionals choose the right tool for structured, constraint-heavy tasks.
In the framework
This term spans Pillar 1 · Foundations and Pillar 2 · Interaction.
Intellectual Property (AI)
Legal frameworks governing ownership of AI-generated content are still evolving. Key questions include who owns AI outputs, whether training on copyrighted material is permissible, and what constitutes original authorship.
Why this matters
Business and marketing students need to understand the IP implications of using AI to create content for professional or commercial use.
In the framework
This term lives in Pillar 4 · Ethics. Explore the full pillar or take the assessment to see where you stand on it.
Interpretability / Explainability (XAI)
The degree to which the internal reasoning of an AI model can be understood and explained to humans. Interpretable AI systems allow stakeholders to understand why a specific output was produced — a core principle of responsible AI.
Why this matters
Explainability is increasingly required by regulation and organizational policy for high-stakes AI decisions in HR, finance, and customer service.
In the framework
This term spans Pillar 1 · Foundations and Pillar 4 · Ethics.
Iteration (Prompt)
The practice of refining a prompt across multiple attempts — adjusting wording, adding context, or changing format — to progressively improve output quality. Rarely does the first prompt produce the best result.
Why this matters
Iteration is the single most underused productivity habit in AI use. Treating prompting as a one-shot interaction leaves significant quality on the table.
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Jagged Technological Frontier
A concept from Harvard Business School research describing how AI is unexpectedly excellent at some tasks and unexpectedly poor at others, with the peaks and valleys difficult to predict in advance.
Why this matters
This mental model is essential for professional judgment about when to trust AI — and it's the research-backed explanation for why AI literacy matters.
In the framework
This term spans Pillar 1 · Foundations and Pillar 3 · Critical Thinking.
Knowledge Cutoff
The date after which an AI model has no training data. Events, research, or developments that occurred after this date are unknown to the model unless explicitly provided in the prompt.
Why this matters
Explains why AI can give outdated information on current events, recent research, or evolving regulations — a critical limitation to communicate to anyone relying on AI for timely information.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Large Language Model (LLM)
A type of AI model trained on massive volumes of text to predict and generate language. ChatGPT, Claude, and Gemini are all LLMs. They produce outputs word-by-word based on statistical patterns, not comprehension.
Why this matters
LLMs are the AI tools students are most likely to use at work — knowing what they are makes you a more effective user.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Machine Learning
A branch of AI in which systems learn patterns from data rather than following hard-coded rules. The model improves through exposure to examples, not explicit programming.
Why this matters
Understanding that AI learns from data — rather than being told answers — explains both its power and its limits.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Meta-prompting
Prompting an AI to help you design better prompts. You ask the model to critique, rewrite, or generate prompt templates for a given task — using AI to improve your AI interactions.
Why this matters
A high-leverage technique for professionals building repeatable AI workflows — particularly in content and strategy functions.
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Mixture of Experts (MoE)
A neural network architecture in which multiple specialized sub-networks are activated selectively for different inputs, rather than using the full model for every query. Used in large models like GPT-4 to improve efficiency.
Why this matters
MoE is the architecture behind several frontier models. Knowing it exists provides context for capability and efficiency comparisons between models.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Model
In AI, a model is a trained system that takes inputs and produces outputs. Models are built through training on data and then deployed for inference. 'The model' and 'the AI' are often used interchangeably in practice.
Why this matters
Understanding that AI tools are specific trained models — not generic intelligence — sets accurate expectations for their capabilities.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Model Card
A brief technical document published alongside an AI model that describes its intended use cases, performance characteristics, limitations, and known biases. A transparency standard for responsible AI development.
Why this matters
Model cards are how developers disclose what their AI can and can't do. Being able to read and interpret one is a practical AI literacy skill.
In the framework
This term spans Pillar 1 · Foundations and Pillar 4 · Ethics.
Model Context Protocol (MCP)
An open standard developed by Anthropic that defines how AI models connect to external tools, data sources, and services — enabling agents to interact with APIs, databases, and applications through a standardized interface.
Why this matters
MCP is rapidly becoming the infrastructure standard for agentic AI integration. Understanding it provides context for how AI agents access and act on real-world data.
In the framework
This term spans Pillar 2 · Interaction and Pillar 7 · Future Readiness.
Model Distillation
A technique in which a smaller, more efficient model is trained to replicate the behavior of a larger, more powerful model — enabling deployment in resource-constrained environments like mobile devices or edge systems.
Why this matters
Distilled models are increasingly used in on-device AI applications. Understanding the trade-offs helps evaluate AI tool choices for practical business use.
In the framework
This term spans Pillar 1 · Foundations and Pillar 7 · Future Readiness.
Model Drift
The degradation in AI output quality or consistency that occurs when a model is updated, replaced, or when the context it was designed for shifts. Workflows built around a specific model version may underperform after an update.
Why this matters
Model drift is a real operational risk for teams with established AI workflows. Knowing to test after updates protects workflow integrity.
In the framework
This term spans Pillar 3 · Critical Thinking and Pillar 7 · Future Readiness.
Model Inversion Attack
A type of adversarial attack in which an attacker uses a model's outputs to reverse-engineer sensitive information from its training data — a real privacy risk when models are trained on personal or confidential datasets.
Why this matters
Understanding that AI models can leak training data shapes how organizations should think about what data they use to train or fine-tune models.
In the framework
This term spans Pillar 1 · Foundations and Pillar 6 · Data & Privacy.
Model Version
AI models are regularly updated, with newer versions typically offering improved capabilities, better accuracy, or new features. Version-specific behavior matters when benchmarking, documenting workflows, or comparing outputs.
Why this matters
Knowing which model version you're using — and that it may change — is important for reproducible professional work.
In the framework
This term spans Pillar 1 · Foundations and Pillar 7 · Future Readiness.
Multi-agent Systems
Networks of AI agents that coordinate with each other to complete complex tasks — one agent plans, another retrieves information, another executes actions — operating with minimal human intervention.
Why this matters
Multi-agent workflows are rapidly moving from research into enterprise deployment. Understanding them prepares students for the next wave of AI-driven automation.
In the framework
This term lives in Pillar 7 · Future Readiness. Explore the full pillar or take the assessment to see where you stand on it.
Multi-turn Context Management
The practice of deliberately structuring a multi-message conversation to maintain coherent context, reinforce instructions, and prevent model drift as the conversation extends and earlier context fades from the window.
Why this matters
Long AI sessions degrade in quality without intentional context management — a practical skill for sustained research or strategy work.
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Multimodal AI
AI systems that can process and generate multiple types of content — text, images, audio, and video — within a single model. GPT-4o and Gemini are examples of multimodal models.
Why this matters
Multimodal tools are rapidly expanding what's possible in marketing, content, and communications roles.
In the framework
This term spans Pillar 5 · Application and Pillar 7 · Future Readiness.
Named Entity Recognition (NER)
An NLP technique that identifies and classifies named elements in text — people, organizations, locations, dates, products — enabling structured data extraction from unstructured content.
Why this matters
NER powers many marketing analytics tools, from media monitoring to competitive intelligence. Understanding it helps students evaluate and configure these tools effectively.
In the framework
This term lives in Pillar 5 · Application. Explore the full pillar or take the assessment to see where you stand on it.
Negative Prompting
Explicitly telling an AI what not to include, do, or produce in its output — for example, 'Do not use bullet points' or 'Avoid jargon.' Negative constraints are as powerful as positive instructions.
Why this matters
Students who only tell AI what to do often get outputs with unwanted elements. Negative prompting is a quick and effective skill upgrade.
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Neural Network
A computational architecture inspired loosely by biological neurons. Layers of interconnected nodes process and transform data, enabling the model to learn complex patterns from examples.
Why this matters
Provides a conceptual mental model for how modern AI processes information — not magic, but math.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
NIST AI Risk Management Framework
A voluntary framework published by the National Institute of Standards and Technology to help organizations identify, assess, and manage AI-related risks across the full AI lifecycle. Increasingly used as a compliance reference.
Why this matters
The NIST AI RMF is the U.S. government's primary AI risk standard. Students entering enterprise or regulated environments will encounter it in AI governance discussions.
In the framework
This term spans Pillar 4 · Ethics and Pillar 6 · Data & Privacy.
On-Device AI / Edge AI
AI models that run directly on a local device — a phone, laptop, or embedded system — rather than sending data to a cloud server. Reduces latency, improves privacy, and enables offline use.
Why this matters
On-device AI is becoming a mainstream product feature. Understanding it explains a growing category of privacy-preserving AI tools.
In the framework
This term spans Pillar 6 · Data & Privacy and Pillar 7 · Future Readiness.
Open-Source AI
AI models whose weights, architecture, and/or training code are publicly released, allowing anyone to run, modify, or fine-tune them. Llama (Meta) and Mistral are prominent examples. Contrasts with proprietary models like GPT-4.
Why this matters
Open-source AI is reshaping the competitive landscape and enabling on-premise deployment for sensitive use cases. Knowing the distinction matters for vendor evaluation.
In the framework
This term spans Pillar 1 · Foundations and Pillar 7 · Future Readiness.
Output Formatting
Instructing an AI to structure its response in a specific way — as a table, numbered list, JSON object, markdown document, or plain paragraph. Explicit format instructions significantly improve output usability.
Why this matters
Unformatted AI outputs often require heavy editing before they're useful. Format instructions are a quick way to reduce downstream effort.
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Overfitting
When an AI model learns the training data so precisely — including its noise and idiosyncrasies — that it performs well on training data but poorly on new, unseen data. A classic model quality failure mode.
Why this matters
Understanding overfitting explains why AI tools sometimes perform impressively in demos but fail in production. It's a key concept for evaluating AI reliability claims.
In the framework
This term spans Pillar 1 · Foundations and Pillar 3 · Critical Thinking.
Parameters
The numerical values within a neural network that are adjusted during training to improve the model's outputs. A model with '70 billion parameters' has 70 billion adjustable values. More parameters generally enable more capability.
Why this matters
Parameter count is often cited in AI model comparisons. Understanding what it means provides context for capability claims.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Persona (Prompt Design)
A detailed description of a role, background, or character assigned to an AI within a prompt — more specific than basic role prompting, a persona can include expertise level, communication style, values, and contextual knowledge.
Why this matters
Well-designed personas dramatically improve the consistency and relevance of AI outputs across repeated use cases like customer communications or brand copy.
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Personalization at Scale
Using AI to generate individually tailored content, recommendations, or communications for large audiences simultaneously — enabling one-to-one marketing at a cost previously impossible without automation.
Why this matters
A transformational capability for marketing professionals — understanding how it works informs both strategy and ethical practice.
In the framework
This term lives in Pillar 5 · Application. Explore the full pillar or take the assessment to see where you stand on it.
Personally Identifiable Information (PII)
Any data that can be used to identify a specific individual — names, email addresses, phone numbers, social security numbers, or combinations of attributes. PII requires special handling when using AI tools.
Why this matters
Entering PII into public AI tools can violate privacy regulations and company data policies.
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
PII Scrubbing
The process of systematically identifying and removing personally identifiable information from text or datasets before they are used with AI tools — often automated but requiring human review for accuracy.
Why this matters
PII scrubbing is a practical workflow step that reduces privacy risk and regulatory exposure when using AI on real business data.
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
Plausibility Bias
The cognitive tendency to accept AI-generated content as accurate because it sounds fluent, professional, and coherent — even without verifying its factual basis. Fluency does not equal accuracy.
Why this matters
The biggest practical risk of using LLMs in professional work. Recognizing this bias is the first step toward building a verification habit.
In the framework
This term lives in Pillar 3 · Critical Thinking. Explore the full pillar or take the assessment to see where you stand on it.
Predictive Analytics (AI-driven)
Using machine learning to analyze historical data and generate probabilistic forecasts — customer churn, purchase intent, campaign performance, demand patterns. Distinct from generative AI, though increasingly combined with it.
Why this matters
Predictive analytics underlies many marketing platforms students will use — segmentation tools, CRM scoring, ad targeting algorithms.
In the framework
This term lives in Pillar 5 · Application. Explore the full pillar or take the assessment to see where you stand on it.
Privacy by Design
A framework in which privacy protections are built into systems and workflows from the beginning — rather than added as an afterthought. One of the foundational principles of GDPR and modern data governance.
Why this matters
Privacy by design is the gold standard for AI tool selection and workflow construction. Students who understand it can advocate for responsible AI practices from day one.
In the framework
This term spans Pillar 4 · Ethics and Pillar 6 · Data & Privacy.
Prompt
The input you give an AI model — a question, instruction, or piece of context that triggers a response. The quality of your prompt directly shapes the quality of the output.
Why this matters
The foundational skill for effective AI use. Better prompts consistently produce better results.
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Prompt Chaining
A technique in which the output of one AI prompt is used as the input for the next, allowing complex multi-step tasks to be completed in a sequence of smaller, more manageable AI interactions.
Why this matters
Prompt chaining is how sophisticated AI workflows are built — the connective tissue between simple prompting and professional-grade AI productivity.
Related terms
- Agent
- Pipeline
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Prompt Engineering
The practice of deliberately crafting inputs — including structure, context, constraints, and examples — to guide AI models toward more accurate, useful, or appropriate outputs. A professional skill, not a technical one.
Why this matters
Prompt engineering is increasingly listed as a required or preferred skill in entry-level job descriptions across marketing and communications.
Related terms
- Few-shot prompting
- Prompt chaining
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Prompt Injection
A security attack in which malicious instructions are hidden in content fed to an AI agent — for example, in a webpage the AI is reading — causing the model to override its original instructions and take harmful actions.
Why this matters
As AI agents are deployed in business workflows, prompt injection becomes a real security and compliance concern. Knowing it exists is the first line of defense.
In the framework
This term spans Pillar 2 · Interaction and Pillar 4 · Ethics and Pillar 6 · Data & Privacy.
Prompt Template
A reusable prompt structure with defined variables and formatting — for example, a template for generating client briefs, weekly reports, or social posts — that produces consistent outputs when applied to different inputs.
Why this matters
Prompt templates are what separate ad-hoc AI use from a repeatable, scalable workflow. Building templates is a core professional skill.
In the framework
This term spans Pillar 2 · Interaction and Pillar 5 · Application.
Prompt Versioning / AI Playbook
The practice of documenting, saving, and iterating on prompt designs and AI workflows — maintaining version history so that changes can be tracked, reviewed, and rolled back when model behavior shifts.
Why this matters
As AI becomes embedded in professional workflows, prompt versioning is the discipline that makes those workflows reliable and auditable.
In the framework
This term spans Pillar 2 · Interaction and Pillar 7 · Future Readiness.
Quantization
A technique for reducing the size and computational requirements of AI models by representing their numerical values with less precision. Quantized models run faster and cheaper but may sacrifice some capability.
Why this matters
Quantized models power many mobile and on-device AI applications. Understanding it explains the capability trade-offs in lightweight AI tools.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Reasoning Model
A class of AI models designed to think through problems step by step before responding — using extended internal reasoning to improve accuracy on complex analytical and logical tasks. OpenAI's o1/o3 and DeepSeek R1 are examples.
Why this matters
Reasoning models represent a meaningful capability shift for business analysis, strategy, and problem-solving tasks — distinct from standard chat models.
In the framework
This term spans Pillar 1 · Foundations and Pillar 7 · Future Readiness.
Responsible AI
A set of principles and practices for developing and using AI systems that are fair, accountable, transparent, and safe. Increasingly codified in organizational policies, government frameworks, and industry standards.
Why this matters
Responsible AI is not just an abstract value — it's a documented expectation in the workplace. Students will be evaluated on whether they use AI responsibly.
In the framework
This term lives in Pillar 4 · Ethics. Explore the full pillar or take the assessment to see where you stand on it.
Retrieval-Augmented Generation (RAG)
A technique in which an AI model retrieves relevant documents or data from an external knowledge base before generating a response, reducing hallucination and improving factual accuracy.
Why this matters
RAG is the architecture behind many enterprise AI tools that need to answer questions about specific company data or current information.
In the framework
This term spans Pillar 1 · Foundations and Pillar 5 · Application and Pillar 7 · Future Readiness.
RLHF (Reinforcement Learning from Human Feedback)
A training technique in which human raters evaluate AI outputs and those preferences are used to further train the model to produce more helpful, accurate, and safe responses. Central to how ChatGPT, Claude, and Gemini were shaped.
Why this matters
RLHF explains why AI models are tuned toward certain values and behaviors — relevant to both capability and ethics discussions.
In the framework
This term spans Pillar 1 · Foundations and Pillar 4 · Ethics.
Role Prompting
Assigning the AI a specific persona or professional role within the prompt — for example, 'Act as a senior marketing strategist' — to shape its perspective, tone, and depth of response.
Why this matters
A simple technique that produces more targeted and contextually relevant outputs for business and communications work.
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Scaling Laws
Empirically observed relationships between model size, training compute, data volume, and performance — showing that AI capabilities tend to improve predictably as these resources scale. Articulated by researchers at OpenAI and Anthropic.
Why this matters
Scaling laws are the empirical foundation behind the rapid growth of AI capabilities. Understanding them provides context for why AI progress has been so fast — and where limits may lie.
Related terms
- Pre-training
- Compute
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Self-Consistency Prompting
A technique in which the same prompt is run multiple times and the outputs are compared to identify the most consistent answer — reducing the impact of randomness and improving reliability for factual or analytical queries.
Why this matters
A practical technique for improving AI output reliability on tasks where accuracy matters more than creativity — research, analysis, and fact-checking.
In the framework
This term spans Pillar 2 · Interaction and Pillar 3 · Critical Thinking.
Sentiment Analysis
A type of natural language processing that classifies the emotional tone of text — positive, negative, or neutral — at scale. Widely used in marketing for social listening, customer feedback analysis, and brand monitoring.
Why this matters
Sentiment analysis is a foundational AI capability in marketing technology. Students using social listening or CRM tools are already using it, often without knowing.
In the framework
This term lives in Pillar 5 · Application. Explore the full pillar or take the assessment to see where you stand on it.
Shadow AI
The use of AI tools within an organization that haven't been officially approved or sanctioned by IT or compliance teams — similar to shadow IT. Common when employees use personal AI accounts for work tasks.
Why this matters
Using unapproved AI tools at work exposes both the employee and the organization to security, privacy, and compliance risks.
In the framework
This term spans Pillar 4 · Ethics and Pillar 6 · Data & Privacy.
Source Attribution
The ability of an AI output to identify or link to the original sources that informed its response. Most LLMs do not reliably attribute sources, which places the verification burden on the user.
Why this matters
Always verify claims independently. Absence of citation in an AI output does not mean the information is unreliable — or reliable.
In the framework
This term lives in Pillar 3 · Critical Thinking. Explore the full pillar or take the assessment to see where you stand on it.
Stochastic Output
AI models are probabilistic — they introduce controlled randomness into their outputs. This means the same prompt will produce different responses each time. This is by design, not a malfunction.
Why this matters
Understanding stochasticity is essential for anyone running AI at scale — reproducibility requires careful prompt design and version control.
In the framework
This term spans Pillar 1 · Foundations and Pillar 3 · Critical Thinking.
Structured Output
AI outputs formatted as machine-readable data structures — JSON, CSV, or XML — rather than natural language prose. Enables downstream automation and integration with other software systems.
Why this matters
Structured outputs are how AI connects to real business systems. Understanding this capability expands what's possible in marketing operations and workflow automation.
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Summarization
Using AI to condense long-form content — reports, articles, meeting notes, research — into shorter summaries. Output quality depends heavily on the quality and length of the source material.
Why this matters
A high-value daily-use skill for business and communications students managing large information volumes.
In the framework
This term spans Pillar 2 · Interaction and Pillar 5 · Application.
Supervised Learning
A type of machine learning in which the model is trained on labeled examples — inputs paired with correct outputs — so it learns to predict outputs for new inputs. Most commercial AI applications use some form of supervised learning.
Why this matters
Explains how AI learns to classify, predict, and make decisions — foundational for understanding how spam filters, recommendation engines, and sentiment analysis work.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Synthetic Data
Artificially generated data that mimics the statistical properties of real data — used to train AI models or test workflows when using real data would pose privacy or security risks.
Why this matters
Synthetic data is the practical solution to the tension between AI development and data privacy. Understanding it enables smarter decisions about when real data is and isn't needed.
In the framework
This term spans Pillar 4 · Ethics and Pillar 6 · Data & Privacy.
Synthetic Media
AI-generated video, audio, or images designed to appear real — including deepfakes, AI voiceovers, and synthetic news anchors. Raises significant authenticity, consent, and misinformation concerns in marketing and communications.
Why this matters
Marketing and communications students will encounter synthetic media both as a creative tool and as a source of potential harm they need to evaluate critically.
In the framework
This term spans Pillar 4 · Ethics and Pillar 5 · Application.
System Prompt
A set of instructions provided to an AI model before the user conversation begins, typically used to define the model's role, tone, constraints, or task context. Users typically do not see system prompts in commercial tools.
Why this matters
Understanding that AI tools are shaped by hidden instructions explains why the same model behaves differently across products.
Related terms
- Role prompting
- Constraints
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Temperature (AI)
A parameter that controls how random or predictable an AI model's outputs are. Higher temperature produces more varied, creative responses; lower temperature produces more consistent, focused ones.
Why this matters
Understanding temperature helps explain why AI outputs vary and how to tune tools for creative versus precise tasks.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Token
The unit AI models use to process text. Tokens are roughly equivalent to word fragments — 'marketing' might be one token, while 'uncharacteristically' might be two or three. Models have token limits for input and output.
Why this matters
Token counts affect cost, response length, and how much content you can include in a single prompt.
Related terms
- Context window
- Embeddings
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Training Data
The dataset used to teach an AI model. The quality, diversity, and recency of training data directly shapes what the model knows, what it gets wrong, and what biases it may carry.
Why this matters
Understanding training data explains why AI can be outdated, biased, or factually wrong about recent events.
In the framework
This term spans Pillar 1 · Foundations and Pillar 4 · Ethics.
Transformer Architecture
The neural network design underlying most modern LLMs. Transformers use an attention mechanism to weigh the relationships between all words in a sequence simultaneously, rather than processing text sequentially. Introduced in the 2017 'Attention is All You Need' paper.
Why this matters
You don't need to build transformers — but knowing what they are gives you a foundation for understanding AI capability and limitation discussions at a professional level.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Transparency (AI)
The principle that AI-generated content, decisions, and processes should be disclosed and explainable to those affected by them — including employers, clients, and audiences.
Why this matters
Workplace AI policies increasingly require disclosure of AI-assisted work. Knowing this protects your professional reputation.
In the framework
This term lives in Pillar 4 · Ethics. Explore the full pillar or take the assessment to see where you stand on it.
Tree of Thought
An advanced prompting technique in which the AI explores multiple reasoning paths simultaneously — branching and evaluating different approaches before converging on an answer. An extension of chain-of-thought for more complex problems.
Why this matters
Tree of thought prompting significantly improves AI performance on multi-step reasoning and strategy problems — relevant for advanced business and analytical applications.
In the framework
This term spans Pillar 2 · Interaction and Pillar 3 · Critical Thinking.
Unsupervised Learning
A type of machine learning in which the model finds patterns in unlabeled data without being told what to look for. Used for clustering, anomaly detection, and dimensionality reduction.
Why this matters
Underpins many analytics and segmentation tools used in marketing — understanding it demystifies how audience segments and customer clusters are built.
In the framework
This term lives in Pillar 1 · Foundations. Explore the full pillar or take the assessment to see where you stand on it.
Vector Database
A database designed to store and retrieve data based on semantic similarity — converting text, images, or other content into numerical vectors and finding the closest matches. Powers search, RAG systems, and recommendation engines.
Why this matters
Vector databases are the infrastructure behind many AI-powered search and personalization tools. Understanding them demystifies how 'semantic search' works.
In the framework
This term spans Pillar 1 · Foundations and Pillar 7 · Future Readiness.
Workflow Orchestration (AI)
The coordination and sequencing of multiple AI tools, models, or agents to complete a complex multi-step task — for example, automatically researching, writing, reviewing, and publishing a content piece.
Why this matters
Workflow orchestration is the practical discipline behind productivity gains from AI. It moves AI use from occasional to systematic.
In the framework
This term spans Pillar 5 · Application and Pillar 7 · Future Readiness.
Zero Data Retention
A setting or agreement in which an AI provider commits not to store, log, or use any inputs or outputs from a user's session. Important for enterprise use cases involving confidential or regulated data.
Why this matters
Understanding data retention options is essential for evaluating whether a given AI tool is appropriate for sensitive work tasks.
In the framework
This term lives in Pillar 6 · Data & Privacy. Explore the full pillar or take the assessment to see where you stand on it.
Zero-shot Prompting
Asking an AI model to complete a task without providing any examples of the desired output. The model relies entirely on its training to interpret and respond to the instruction.
Why this matters
Default interaction mode for most users — knowing its limitations helps you know when to add examples.
In the framework
This term lives in Pillar 2 · Interaction. Explore the full pillar or take the assessment to see where you stand on it.
Knowing the words is a start. Knowing how to use them is the assessment.
See where you stand on all 7 pillars in about 10 minutes, and find out which terms here are skills you actually use.