AI, explained
What is an LLM (large language model)?
It is a prediction engine for language: trained on enormous amounts of text to guess what comes next, which turns out to be enough to write, summarise, translate, and reason.
Strip away the mystique and a large language model does one thing: given some text, it predicts what should come next, one token at a time. What is surprising is how far that goes. Train that prediction on a large enough slice of human writing, and the model absorbs grammar, facts, styles, reasoning patterns, and code. Ask it to continue "the capital of Portugal is" and it produces "Lisbon", not because it looked it up, but because that is the overwhelmingly likely continuation it learned.
What an LLM is good at
- Language work. Drafting, rewriting, summarising, and translating, including across Portuguese, English, and French.
- Following instructions. Turning a plain-language request into structured output, a classification, or a formatted answer.
- Reasoning over provided text. Answering questions about a document you give it, extracting the key points, comparing options.
What it is not
An LLM is not a database and not a search engine. Its knowledge is frozen at training time and stored as patterns, not facts it can cite, which is why it can state something false with complete confidence. This behaviour is called hallucination, and it is inherent, not a bug that gets patched away. It also has no live access to your business unless you connect it, and no memory of yesterday's conversation unless you provide it.
How businesses make it reliable
The gap between an impressive demo and a dependable tool is closed with engineering around the model, not the model alone. Retrieval (RAG) grounds answers in your real documents so they are current and citable. Tools let the model take actions in your systems under controlled permissions. Guardrails and human review handle the cases where being wrong is costly. The model is the engine; the reliability comes from the car built around it.
Why the plumbing matters more than the pick
Companies often fixate on which model to use, but for most business problems several leading models are more than capable, and the winning factor is how well the surrounding system feeds them the right context and constrains their output. Choosing a model is a decision you should be able to reverse cheaply, which is one reason building against standards rather than a single provider pays off. If you are weighing where an LLM would actually earn its keep in your operation, an AI audit is the practical starting point.
Frequently asked questions
What is an LLM?
An LLM, or large language model, is an AI trained on huge amounts of text to predict the next piece of language given what precedes it. At scale, that objective produces a system that can write, summarise, translate, answer questions, and follow instructions. LLMs power assistants such as ChatGPT and Claude and are the core of most modern business AI.
What is the difference between an LLM and ChatGPT?
An LLM is the underlying model; ChatGPT is a product built on one. ChatGPT wraps a large language model in a chat interface, safety layers, memory features, and tools. The model provides the language ability, while the product decides how you interact with it. The same distinction applies to Claude, Gemini, and other assistants.
Can an LLM be wrong?
Yes, and confidently so. An LLM generates the most likely continuation of text, not verified fact, so it can produce fluent but false statements, known as hallucination. Its knowledge is also frozen at training time. Businesses reduce this by grounding answers in real documents through retrieval and adding human review for high-stakes outputs.
Related
Ready to put AI to work?
Book a discovery audit and we will map the highest-ROI AI agents and automations for your business.
Book a discovery audit →