AI glossary for leaders
The Jargon Decoder from inside the certifications, open to everyone.
34 terms in plain English, the words a leader actually meets when the room turns to AI. Each links to the module it first comes up in. Members get the same decoder one tap away inside every lesson.
- Agent
- AI that takes multi-step actions toward a goal you set (searching, writing files, using tools) instead of just answering a question. More powerful, and more in need of guardrails.
- API
- The plug through which your software calls someone else's AI. 'We use the API' means renting the model per use rather than owning infrastructure.
- Automation
- Software doing a task end-to-end without a person. AI widened what's automatable; the leadership question is which steps still deserve human judgment.
- Context window
- How much material an AI can consider at once. Think of it as working memory. Bigger windows mean it can read your whole report; nothing outside the window exists for it.
- Copilot
- AI embedded inside a tool people already use (email, docs, CRM) as an assistant. The gentlest on-ramp for teams, and the easiest spend to lose track of.
- Data readiness
- Whether your data is accessible, clean, and permitted for the AI use you're planning. The unglamorous predictor of most AI project outcomes.
- Deepfake
- AI-generated audio or video that convincingly imitates a real person. Good enough now that a familiar face or voice on a call is no longer verification; a callback on a channel you chose still is.
- Deployment
- A tool put into real daily use, as opposed to a pilot. The pilot is the test with a few people and an exit. The deployment is the thing your staff and customers now depend on, which is why it needs owners, a review step, and a way to switch it off.
- EU AI Act
- The European Union's AI law, phasing in from 2025 through 2028. It sorts AI uses into risk tiers with duties for each; anyone selling into or operating in the EU needs a view on it.
- Fine-tuning
- Training an existing AI model further on your own examples so it behaves more like you need. Expensive and usually unnecessary: most companies get further with better prompts and context.
- GenAI
- Generative AI: models that produce new text, images, code, or audio rather than just classifying existing data. The wave that made AI a board topic.
- Guardrails
- The rules and controls that keep AI use safe: what data may go in, what decisions need a human, what the AI may never do. Guardrails are what let you say yes quickly.
- Hallucination
- When an AI states something false with total confidence. It is not a glitch. It is built into how these models work. The fix is process (verification, human review), not scolding the tool.
- Human-in-the-loop
- A workflow where AI drafts and a person approves before anything ships. The single most useful risk control you have: speed of AI, accountability of a human.
- Inference
- Running a trained model to get answers (as opposed to training it). Inference cost is the per-use bill, the number that scales with adoption.
- ISO 42001
- The international management-system standard for AI (ISO/IEC 42001): a documented cycle of policy, risk assessment, controls, and review. Certification is an audit you can buy; alignment is a truthful mapping of governance you already run.
- LLM
- Large Language Model: the engine behind tools like ChatGPT and Claude. It predicts the most likely next words, which makes it brilliant at drafting and summarizing, and capable of confidently making things up.
- Machine learning
- Software that learns patterns from data instead of following hand-written rules. The umbrella over almost everything called AI today.
- Model
- A trained AI system, the product of feeding enormous data through training. Different models trade off capability, speed, and price; picking one is a procurement decision, not a loyalty oath.
- Model risk
- The chance an AI system is wrong in ways that hurt: biased decisions, fabricated facts, drifting quality. Manage it like any other operational risk, with measurement, an owner, and a review cycle.
- NIST AI RMF
- The US National Institute of Standards and Technology's AI Risk Management Framework: free, voluntary, and organized into four functions, govern, map, measure, manage. Increasingly used as procurement shorthand, and in Texas as a statutory safe harbor.
- Open-source model
- A model whose weights you can download and run yourself. More control and privacy, more work. The alternative to renting intelligence through an API.
- PII
- Personally Identifiable Information: anything that can identify a person. The data category that decides whether an AI use needs privacy review before it needs enthusiasm.
- Pilot
- A deliberately small, time-boxed test of an AI use case with success criteria set in advance. The alternative to both paralysis and company-wide rollouts of the untested.
- Production
- The live setting where real work happens, as opposed to a trial. "In production" means real customers, real money, and real consequences for a mistake. It has nothing to do with a factory or manufacturing output.
- Prompt
- The instruction you give an AI. The quality of what you get back tracks the quality of the brief you give, exactly like delegating to a smart new hire.
- RAG
- Retrieval-Augmented Generation: letting an AI look things up in your documents before answering, so responses use your facts instead of its memory. The standard fix for 'the AI doesn't know our business.'
- ROI
- Return on investment. For AI: hours saved, revenue gained, or risk reduced, minus what it costs to run. Measured, not asserted in a vendor deck.
- shadow AI
- Employees using AI tools nobody approved, with company data. Universal, mostly well-intentioned, and the first thing an AI use policy has to make safe rather than pretend to ban.
- Token
- The unit AI vendors bill by, roughly three-quarters of a word. 'Cost per token' is why long documents and chatty automations have real, countable costs.
- Training data
- Everything a model learned from. It sets the model's knowledge cutoff and its blind spots. It is also why vendors get asked 'was our data used to train this?'
- Use case
- One specific job AI does for the business: 'summarize support tickets,' not 'do AI.' Strategy is choosing a few high-value use cases and finishing them.
- Vendor lock-in
- When leaving a supplier becomes too expensive because your data, workflows, or prompts are trapped in their product. Avoided by asking exit questions before signing.
- Workslop
- AI-generated work output polished enough to pass along and empty enough to push the real work onto whoever receives it. Named in 2025 research; the cost lands on the receiver, and the reputational cost lands on the sender.
Where this comes up: Certified AI Leader, Thinking in AI
Where this comes up: Certified AI Leader, The AI Landscape for Business
Where this comes up: Certified AI Leader, The AI Landscape for Business
Where this comes up: Certified AI Leader, The AI Landscape for Business
Where this comes up: Certified AI Leader, AI Strategy Foundations
Where this comes up: Certified AI Leader, The AI Landscape for Business
Where this comes up: Certified AI Leader, Guardrails: Ethics, Risk & Responsibility
Where this comes up: Certified AI Leader, AI Strategy Foundations
Where this comes up: Certified AI Leader, Thinking in AI
Where this comes up: Certified AI Leader, Thinking in AI
Where this comes up: Certified AI Leader, Thinking in AI
Where this comes up: Certified Chief AI Leader, Governance, Ethics & Risk at Scale
Where this comes up: Certified AI Leader, Working with AI as a Leader
Where this comes up: Certified AI Leader, The AI Landscape for Business
Where this comes up: Certified AI Leader, Thinking in AI
Where this comes up: Certified Chief AI Leader, Governance, Ethics & Risk at Scale
Where this comes up: Certified Chief AI Leader, Governance, Ethics & Risk at Scale
Where this comes up: Certified AI Leader, AI Across the Business
Where this comes up: Certified AI Leader, The AI Landscape for Business
Where this comes up: Certified AI Leader, The AI Landscape for Business
Where this comes up: Certified AI Leader, The AI Landscape for Business
Where this comes up: Certified AI Leader, Thinking in AI
Where this comes up: Certified AI Leader, AI Strategy Foundations
Where this comes up: Certified AI Leader, The AI Landscape for Business
Where this comes up: Certified AI Leader, Thinking in AI
Where this comes up: Certified AI Leader, AI Strategy Foundations
Where this comes up: Certified AI Leader, Thinking in AI
Where this comes up: Certified AI Leader, AI Across the Business