The Marsen glossary
Large Language Model
A machine-learning model trained to process and generate language by predicting useful sequences from context.
The short answer
A large language model (LLM) is a machine-learning model trained on large amounts of text to predict and generate language. It can summarise, classify, draft and answer questions from the context it is given, but fluent output is not the same as accurate output, so business workflows supply it with approved information and check important results.
How an LLM works
During training, the model learns statistical patterns in text: which words and ideas tend to follow others. At use time, it receives a prompt (instructions plus any context you supply, such as a customer email or a policy document) and generates a response one token at a time. It has no live access to your systems unless a workflow gives it tools, and its built-in knowledge stops at its training cut-off. That is why business uses pair an LLM with retrieval of current documents, structured outputs and rules about what it may do.
A worked example
A support team receives long customer emails. An LLM in an n8n workflow reads each email and returns a structured result: the customer's issue in one sentence, the product mentioned, the urgency, and a suggested reply drawn from the current returns policy that the workflow retrieves and includes in the prompt. A person reviews the suggested reply before it is sent. The model saves reading time; the policy document and the reviewer keep the answer correct.
- Input: customer email plus the current returns policy
- Output: one-sentence summary, product, urgency, draft reply
- Control: a person approves the reply before sending
What to check before using one
Give the model the information it needs rather than relying on what it learned in training. Ask for structured output that a workflow can validate. Test with real examples, including ambiguous and adversarial ones. Decide which outputs can be used automatically and which need review, and keep personal or confidential data out of providers your agreements do not cover.
See how retrieval-augmented generation helps
Design human approval for AI actions
Common misunderstandings
An LLM does not look facts up unless a tool or retrieval step provides them. A confident answer can still be wrong, and the same prompt can produce different answers on different runs. Larger models are not always better for a narrow task; a smaller model with good context and validation is often cheaper and more predictable.
Take the next step
Explore the related guide, or look up another term.
Built around your business
Put what you’ve learned to work.
Book a demoBring us your questions. We’ll help you find a useful next step for your business.