Prompt Builder versus Agent Builder
Key Takeaway
Use Prompt Builder when generative AI belongs inside work users already do, such as a Field Generation prompt template that fills a field on a Lightning record page. Use Agentforce Builder when people need a conversational agent that chooses its own actions. When an agent action needs output grounded in your CRM data, the action calls a prompt template built in Prompt Builder.
Prompt Builder and Agentforce Builder are two separate tools for adding generative AI to Salesforce. Prompt Builder creates reusable prompt templates that ground an LLM in your CRM data, so users get AI output inside their normal work, like a drafted email or a filled-in field. Agentforce Builder builds agents that hold conversations, decide which actions to take, and complete tasks. The tools also work together, because an agent can call a prompt template through an action. To pick the right tool for a requirement, look at how people interact with the AI.
What Prompt Builder does
Prompt Builder is where you create, test, revise, customize, and manage prompt templates. A template pulls in CRM data through merge fields. Merge fields can reference record fields, flows, related lists, and Apex.
Prompt Builder adds generative-AI moments to users' daily workflow. Prompt template types include these:
| Template type | What it does |
|---|---|
| Sales Email | Drafts personalized emails for a contact or lead |
| Field Generation | Lets users fill a field on a Lightning record page with an LLM-written summary or description |
| Flex | Takes a variable number of inputs of different object types, for example to generate a newsletter from product and campaign objects |
Prompt templates are not tied to one place. You can call a template through an invocable action anywhere on the Salesforce Platform. You can also call a template through the Connect REST API or Connect in Apex.
Assign the "Prompt Template Manager" permission set to Salesforce admins who create and manage prompt templates in Prompt Builder.
A sales director wants each opportunity to show a short, plain summary of where the deal stands. The admin creates a Field Generation prompt template for a custom "Deal Summary" field on the Opportunity object. Merge fields in the template pull in the opportunity's stage, amount, and next step. When a rep opens an opportunity, the rep fills the Deal Summary field with an LLM-written summary without leaving the record page. No agent is involved, because the AI output sits inside a page the rep already uses.