Prompt Engineering20% of the exam
Model access controls and bring your own LLM
How model use is governed: the model chosen per prompt template version, org-level provider blocking in Setup, hiding an LLM configuration in AI Models (formerly Einstein Studio), beta-model enablement, limiting who holds Manage Prompt Templates, Salesforce's AI policies, and connecting a contracted provider with bring your own LLM.
Explain how to manage and prevent specific models from being accessed.
~7 min read · 6 practice questions ·
What the exam expects
- Pin a prompt template to an approved model by selecting it in the template, and explain why prompt text can't change routing.
- Prevent a model from being used anywhere in the org: turn off its provider in Setup or hide its LLM configuration in AI Models, and know what each does to existing templates.
- Govern who can choose models by limiting the Manage Prompt Templates permission, alongside Salesforce's AI Acceptable Use Policy and model containment policies.
- Connect a contracted provider account with bring your own LLM (BYOLLM) and use it from Prompt Builder while keeping the Trust Layer.
- Key terms: Salesforce-managed model (a model Salesforce provides out of the box, enabled by default); AI Models (formerly Einstein Studio: where model configurations are created, tested and shown or hidden); LLM configuration (a configured model that Prompt Builder can select); BYOLLM (connecting a foundation model through your own provider account); model provider access (the org setting that allows or blocks a provider's models).
Key facts
Default model availability
- Detail
- Salesforce-managed models are enabled by default
Block a provider
- Detail
- Admin turns off a model provider in Setup; its models are blocked in the org
Provider flags
- Detail
disableAIProvOpenAI,disableAIProvAzureOpenAI,disableAIProvAWSBedrock,disableAIProvVertexGemini; default false
Hidden configuration
- Detail
- Not selectable in Prompt Builder; templates already using it run as before
Beta models
- Detail
- Off by default; enable in Einstein Setup, preferably in sandboxes
Where a template's model lives
- Detail
primaryModelon each prompt template version
Agent default model
- Detail
- Changed on the Agentforce Agents page in Setup
BYOLLM providers
- Detail
- Azure, Bedrock, OpenAI or Vertex accounts; others via LLM Open Connector
BYOLLM and the Trust Layer
- Detail
- Inference by the provider; Trust Layer features fully supported
Who can change a template's model
- Detail
- Holders of Prompt Template Manager or Customize Application
Sources for this section (6)
- Large Language Model Supporthelp.salesforce.com (opens in a new tab)
- EinsteinGptSettingsdeveloper.salesforce.com (opens in a new tab)
- Hiding Configurationshelp.salesforce.com (opens in a new tab)
- GenAiPromptTemplatedeveloper.salesforce.com (opens in a new tab)
- Supported Modelsdeveloper.salesforce.com (opens in a new tab)
- Enable Prompt Builderhelp.salesforce.com (opens in a new tab)
- Why doesn't hiding a model configuration break existing prompt templates?
- When should you turn off a model provider rather than hide a configuration?
- Why does BYOLLM still meet Trust Layer requirements?
- What do model containment policies do?
The questions are free to read. Answers with their sources and self-grading are included with the pack.
Keep reading with the Agentforce Specialist pack
- Core concepts
- Exam traps
- Worked scenario
The pack includes:
- All 26 lessons of the Agentforce Specialist guide, with reading progress
- Interactive decision walks and step-by-step sequence steppers
- Inline practice questions and one-click topic practice (6 questions for this lesson)
- An official source cited for every claim
- An AI tutor that answers from the guide
- Unlimited timed practice exams