AI Model
An AI Model in Salesforce is a machine learning or large language model registered and configured for use within the Einstein 1 Platform, the AI layer that powers Agentforce.
Definition
An AI Model in Salesforce is a machine learning or large language model registered and configured for use within the Einstein 1 Platform, the AI layer that powers Agentforce. AI Models can be Salesforce-hosted foundation models (the capability formerly marketed as Einstein GPT), third-party models from providers like OpenAI, Anthropic, or Google accessed through Salesforce's LLM Gateway, or custom models brought in via Bring Your Own Model (BYOM). They are managed in Einstein Studio (Model Builder) and invoked by prompt templates, Agentforce agent actions, and Einstein predictions.
In plain English
“Think of an AI Model as the 'brain' that Salesforce uses to answer questions or make predictions. Just like you might ask a smart friend for advice, Salesforce asks one of these brains, and the brain uses what it learned from lots of examples to give an answer. Some brains are built by Salesforce, some come from other companies, and some you can bring yourself.”
Worked example
Cypressbrook Capital's Einstein 1 Platform setup uses three different AI Models for different tasks: a Salesforce-hosted foundation model for standard prompt-template completions, an Anthropic Claude model accessed through the LLM Gateway for high-quality long-form drafting, and a custom fraud-detection model the data-science team trained and registered via Bring Your Own Model (BYOM). Einstein Studio's Model Builder is where each is registered and tuned; prompt templates and Agentforce actions invoke them by name. The team picks the right AI Model per use case rather than forcing every workload onto one.
Why AI Model matters
AI Models are registered and managed in Einstein Studio, which acts as the control panel for any model that Salesforce apps will call. Registration captures the model endpoint, authentication, and schema, so that prompt templates, agent actions, and Einstein predictions can invoke the model consistently without each builder needing to know the underlying wiring.
The Bring Your Own Model (BYOM) capability lets enterprises use models they already host on Amazon SageMaker, Google Vertex AI, or Databricks without moving data out of those environments. Combined with the Einstein Trust Layer, this means prompts and responses can flow through masking, toxicity detection, and audit logging regardless of which provider's model is doing the work.
How organizations use AI Model
Registered a custom fraud-detection model hosted on SageMaker as an AI Model in Einstein Studio, then invoked it from a Flow on the Case object. Every incoming high-value claim is now scored in real time, with the model's confidence written back to a custom field that routes suspicious cases to a human reviewer.
Uses a third-party OpenAI model as an AI Model for generating personalized trip itineraries inside a prompt template. The marketing team iterates on the prompt without touching the model configuration, and swapping to a newer model version takes one setting change rather than a redeployment.
Brought their existing risk-scoring model into Salesforce via BYOM so that the same model the finance team uses in their data warehouse now powers opportunity scoring in Sales Cloud. This eliminated the drift between what finance and sales saw as 'high-risk' deals.
Trust & references
Straight from the source - Salesforce's reference material on AI Model.
- Connect a Vertex AI ModelSalesforce Help
- Use Google Cloud Vertex AI ModelsSalesforce Help
Test your knowledge
Q1. Where are AI Models registered and managed in Salesforce?
Q2. What does Bring Your Own Model (BYOM) allow?
Q3. Which of these is NOT a typical consumer of an AI Model in Salesforce?
Discussion
Loading discussion…