What it is
An LLM is a type of AI trained on enormous amounts of text — books, websites, business documents — until it learns to understand language and generate coherent, context-aware responses. When you ask Agentforce to summarize a case, draft an email, or decide next steps, an LLM is doing the heavy lifting. It does not just match keywords; it reads intent, follows nuanced instructions, and produces output that sounds like a knowledgeable human wrote it.
Why it matters
LLMs are why AI suddenly got useful for business. They can understand context, follow instructions, write content, analyze documents, and reason through problems. Salesforce uses LLMs (including partnerships with OpenAI and Anthropic) to power Agentforce.
How it connects
Salesforce does not build its own LLM from scratch — it partners with providers like OpenAI and Anthropic and layers its own data, security, and business logic on top through the Einstein Trust Layer, so the model works with your Salesforce data without exposing it.
Key players
OpenAI (GPT-4), Anthropic (Claude), Google (Gemini), Meta (Llama)
Good to know
The LLM is powerful but only as useful as the instructions and context it receives — vague prompts or missing CRM data produce generic answers, which is why prompt design and clean Salesforce data matter more than most teams expect.
Related terms
RAG (Retrieval-Augmented Generation)
A technique that makes AI smarter by fetching relevant information from your data before generating a response. The AI "looks it up" instead of guessing.
AI Agent
An autonomous AI system that can perceive its environment, make decisions, and take actions to achieve specific goals - without constant human direction.
Prompt Builder
Salesforce's tool for creating reusable, templatized prompts that pull in live CRM data. Build once, use across flows, agents, and record pages.