Large Language Model (LLM)
The engine behind tools like ChatGPT and Claude. It predicts text — which turns out to be enough to draft, summarize, analyze, and converse.
Why it matters: think of it as a brilliant, fast, occasionally unreliable new hire who's read most of the internet but knows nothing about your business yet.
Tokens
How AI reads and bills. A token is a chunk of text — roughly three-quarters of a word. Models charge per token in and out.
Why it matters: length and cost are the same thing. Long documents and rambling prompts are a line item.
Context Window
The model's short-term memory — how much it can hold in view at once. Go past it and earlier details quietly fall away.
Why it matters: it's why an AI loses the thread on a long document, and why how you feed it information changes the result.
Prompt
The instruction you give the model. Output quality is mostly downstream of input quality.
Why it matters: most disappointing AI results are actually disappointing prompts. "Prompting" is a real, learnable skill.
Hallucination
When the model states something false with total confidence. It isn't lying — it's predicting plausible text.
Why it matters: never put AI on anything high-stakes without a human check or a grounding method (see RAG).
RAG (Retrieval-Augmented Generation)
Giving the model your actual documents to answer from, instead of relying on its general training.
Why it matters: this is how you make AI accurate about your business — your policies, your data — and sharply cut hallucinations.
Fine-tuning
Training a base model further on your own examples so it adopts your style, format, or domain.
Why it matters: powerful, but often overkill. Good prompting and RAG solve most problems first, at a fraction of the cost.
Agent
AI that doesn't just answer but acts — taking steps, using tools, completing multi-step tasks on your behalf.
Why it matters: this is where 2026's real leverage lives — the shift from AI that describes work to AI that does it.
MCP (Model Context Protocol)
An emerging standard that lets AI plug into your tools — calendar, email, databases, apps — through a common connector.
Why it matters: it's quietly becoming the wiring that turns a chatbot into something that operates inside your business.
The Learning Loop
The system where your work continuously makes your AI better, and that better AI makes your work faster — on knowledge you own.
Why it matters: the model is rented and commoditizing fast. The loop is yours and compounds. It's the whole game — and the thing we build.
