LLM (Large Language Model)

A large language model is the underlying system trained on text to generate and reason about language. It is the model, not the platform built around it.

An LLM takes text and produces text. Everything else people associate with AI products, remembering yesterday, reaching a CRM, respecting permissions, stopping for approval, belongs to the system around the model. Confusing the two leads to a common disappointment: a company adopts a very capable model and finds it still knows nothing about the company.

This is also why platform choice and model choice are separable decisions. Models improve on a fast cycle and can be swapped. The connected context, permissions and workflows around them are the part that takes time to build and the part that determines whether the model is useful at work.

Frequently Asked Questions

No. The assistant is the product: an interface, memory, connections to tools and rules about what it may do. The LLM is the component that generates the language.

Usually not. If the system never retrieved the right context, a stronger model produces a more fluent version of the same wrong answer.

Not inherently. Supplying context at request time is not training. Whether anything is retained for training is a contractual and configuration question worth confirming explicitly.