[GTM glossary](/glossary.md) Large language model Contextual Intelligence & AI # Large language model (LLM) In short A large language model (LLM) is an AI model trained on very large text collections to predict the next word in a sequence. That simple skill lets it write, summarize, classify, translate, and answer questions in plain language. For go-to-market teams, an LLM is the engine behind most modern sales and marketing AI. It drafts emails, summarizes calls, reads websites, and turns messy text into structured fields. It does not know your accounts unless you give it that information. ## How an LLM works in plain English An LLM breaks text into small pieces called tokens and learns statistical patterns between them. When you send a prompt, it generates a response one token at a time, choosing likely continuations. Because it predicts text rather than looking up facts, it can sound confident and still be wrong. This is why teams pair LLMs with trusted data and review steps. --- ## Where GTM teams use LLMs * **Research:** Summarize a company website, news, or job postings into a short account brief. * **Writing:** Draft first versions of emails, call scripts, and landing page copy for a human to edit. * **Classification:** Tag companies by industry or use case from free text, or label inbound replies as interested, not now, or unsubscribe. * **Extraction:** Pull names, tools, and pain points out of call transcripts and notes. --- ## Limits to plan for * **Hallucination:** The model may invent details. Ground it in real records and check anything that will be sent to a prospect. * **Stale knowledge:** A model only knows what it saw in training unless you pass in fresh information. * **Context window:** It can only read a limited amount of text per request, so long inputs must be trimmed or retrieved selectively. * **Privacy:** Check what data leaves your systems before you paste customer information into any tool. --- ## How to get useful output ### Give it context Say who the audience is, what the goal is, and include the facts it should use. Say a rep asks for an email to a VP of Operations. Adding two verified facts about the account gives a far better draft than the title alone. ### Constrain the format Ask for a specific length, tone, and structure. If the output feeds another system, request a fixed set of fields. ### Keep a human in the loop Use the model to speed up the first draft, not to replace judgment on who to contact and what to promise. --- ## Large language model FAQs **What is a large language model (LLM) in simple terms?** A large language model is an AI system trained on vast amounts of text so it can read and write natural language. It predicts likely next words, which lets it answer questions, summarize documents, and draft content in a conversational way. **How do sales teams use a large language model?** Sales teams use a large language model to summarize account research, draft outreach, classify replies, and extract details from call transcripts. Output is best treated as a first draft that a rep checks before anything reaches a prospect. **Why do large language models make things up?** A large language model generates text by predicting likely words, not by checking a database of facts. When it lacks reliable information it can produce plausible but false statements. Supplying verified source data and reviewing output reduces the risk. **Is an LLM the same as machine learning?** An LLM is a specific kind of machine learning model built for language. Machine learning is the broader field that also covers models for scoring, forecasting, and image recognition. All LLMs are machine learning models, but most machine learning models are not LLMs. Share term --- ## More related phrases * [Machine learning](/glossary/machine-learning.md) * [Generative engine optimization](/glossary/generative-engine-optimization.md) * [LLM optimization](/glossary/llm-optimization.md) ## Similar phrases * [Machine learning model](/glossary/machine-learning-model.md) A machine learning model is a program that learns patterns from historical data to make predictions or decisions on new data, such as which accounts are likely to convert. * [AI hallucination](/glossary/ai-hallucination.md) An AI hallucination is output from a language model that sounds confident and plausible but is false, invented, or unsupported by any real source, a real risk in sales research and outreach. * [Prompt engineering](/glossary/prompt-engineering.md) Prompt engineering is the practice of writing and refining instructions for AI models so they return accurate, consistent, and useful output for tasks such as research, email drafting, and account summaries. * [AI for sales](/glossary/ai-for-sales.md) AI for sales means using machine learning and generative AI to help sellers find accounts, prepare, write, follow up, and forecast, while humans run the conversations and decisions. * [Context engineering](/glossary/context-engineering.md) Context engineering is the practice of deciding what information an AI model receives, in what form and order, so it can complete a task accurately, such as supplying verified account data to a sales agent.