Glossary

The terms you need — without the jargon

Eighteen AI terms, defined the way an executive actually needs them: what the thing is, why it matters, and what it costs you to get wrong. No prerequisite reading required.

01

The terms

LLM (Large Language Model)
A software model trained on huge amounts of text that predicts the next word in a sequence. It’s the engine behind ChatGPT and similar tools. It doesn’t “know” facts — it recognizes patterns, which is why it can sound right while being wrong.
RAG (Retrieval-Augmented Generation)
A technique that lets an AI read your documents before answering. Instead of relying only on what it was trained on, it pulls the relevant pages from your files and grounds its answer in them — closer to an open-book exam than recall from memory.
Agentic AI
AI that takes multi-step actions toward a goal, not just answers one question. It can break a task into steps, use tools (search, email, your systems), and decide what to do next. More powerful — and with more moving parts to govern.
Fine-tuning
Retraining a model further on your specific data so it gets better at a narrow task or your house style. Useful when a general model isn’t specific enough — and overkill when good prompting or RAG would already do the job.
Prompt engineering
Writing instructions that reliably get useful output from a model. Part craft, part discipline: clear goals, context, examples, and constraints. It’s the cheapest lever to pull before you buy anything new.
AI operating leverage
Getting more output from the same headcount by embedding AI into how work actually gets done. The goal isn’t a chatbot on the website; it’s a measurable change in capacity and cycle time across real workflows.
Operating Partner
Quiet Turn’s staffed engagement: the Principal plus a small team who carry AI work between sessions, for the executive who wants it done and not just learned. Senior AI judgment on demand, without hiring a full-time AI executive.
Decision rights
Who is allowed to make which call, and how. With AI: which decisions a person still owns, which an AI may draft, and which an AI may act on — and who signs off. It’s the governance question hiding under the technology one.
Operating model
How your organization actually turns inputs into value: roles, processes, handoffs, and decision flows. AI changes this, not just the tools — and that’s where most of the return (and most of the risk) actually lives.
Capacity vs. productivity
Capacity is how much your team can take on; productivity is output per unit of input. AI often raises capacity first. The strategic question is whether you convert that into growth, lower cost, or freed-up talent — or it just quietly evaporates.
AI governance
The rules, roles, and review processes that keep AI use safe, compliant, and aligned with how you want to operate. Covers data handling, model oversight, human checkpoints, and board-level accountability.
Model context window
How much text a model can consider at once — its “working memory.” Affects how much of your documents it can read in a single pass. Bigger isn’t always better: it costs more and can dilute the model’s focus.
Hallucination
When a model states something false with full confidence. It isn’t lying; it’s pattern-matching without verification. This is why AI output for any consequential decision needs a human in the loop.
Inference
Running a trained model to produce an answer — the moment of use, as opposed to training. Most of the ongoing cost and latency of an AI system comes from inference, not the one-time training.
Embedding
Turning text (or images) into numbers that capture meaning, so similar things sit close together in “space.” The foundation of search, clustering, and RAG — it lets software find things by meaning, not just by keyword.
Generative AI
AI that produces new content — text, images, code — rather than just classifying or labeling what exists. ChatGPT is generative; a spam filter is not. Most of the current executive buzz is about this category.
Tokens
The chunks (roughly word-parts) a model counts as it reads and writes. Pricing, context windows, and speed are all measured in tokens. A useful unit for budgeting what AI usage actually costs.
Vector database
Storage optimized for embeddings, letting you search “find the documents most similar in meaning to this one.” It’s the plumbing behind RAG and semantic search — important to know exists, rarely something you buy yourself.
02

Use these in a real decision

The glossary is the vocabulary. Private Executive AI Coaching is where it meets your operating model — and where you find the one to three moments worth testing.

Book the Operating Leverage Session

Four coaching engagements run at a time.

See how this compares to hiring or an agency →

Operating Leverage Session