TENCONIC / AI GLOSSARY

AI glossary

From unfamiliar terms
to the differences that matter at work.

Understand the concept first, then explore your next question through related terms.

20 terms

AI fundamentals

Understand how models read information and produce responses.

What is a large language model (LLM)?
How does AI write sentences, and why can it be wrong?
What is a token?
Why are character counts and token counts different?
What is context in AI?
What information does AI use when answering?
What is a context window?
Does reading a long document mean remembering earlier conversations?
What is context engineering?
What should change when rewriting a prompt is not enough?

Explore retrieval, current information, and traceable evidence.

What is retrieval-augmented generation (RAG)?
Does connecting company documents mean retraining the model?
What is an embedding?
What changes when documents are represented as numbers?
What is vector search?
Can search find related documents without matching words?
What is enterprise context?
Why can the same question require different answers at different companies?
What is data freshness?
Why does AI still answer with the old policy after an update?
What is provenance in AI?
Can you trace an AI answer back to the information behind it?

Distinguish persistent information from model training.

What is AI memory?
Is saving a conversation the same as AI remembering it?
What is long-term memory in AI?
What information can continue into a new conversation?
What is the memory lifecycle?
How should incorrect or outdated memories be handled?
What is AI personalization?
What can be personalized beyond using someone’s name?
How do fine-tuning, in-context learning, and continual learning differ?
Does saying an AI learns always mean the model itself changes?

Learn how work continues and how to evaluate it.

What is an AI agent?
How does answering a question differ from carrying out work?
What is a stateful agent?
Can an interrupted AI task continue the next day?
What are AI evaluation and a proof of concept (PoC)?
How does a good demonstration differ from buying value?
What are shared memory and data isolation?
How can teams share memory while keeping customer data separate?