Chapter 3
Copilot, Agentic AI and LLMs
Preview published October 21, 2025
Summary
The data science behind AI: how training data is gathered and refined, how tokens turn text into numbers, why an LLM predicts rather than understands, and how that explains both the power and the failures of Copilots and agentic AI.
Additional explanation
Key terms from this chapter: token, tokenization, training corpus, model collapse, data provenance.
LLMs predict the next token based on probability; there is no comprehension behind the output. Understanding that is the foundation for managing accuracy and risk in Copilot deployments.