Section 2 · Prompt Engineering · 20% of exam
2.1 When to Use Prompt Builder
Exam objective
Given a business requirement, identify when Prompt Builder is an appropriate way to produce generative AI output in Salesforce.
Prompt Builder is Salesforce’s low-code workspace for creating, testing, and managing reusable prompts that call a large language model (LLM). Use it when a Salesforce process needs generated language, such as a record summary, a draft email, a response grounded in CRM data, or a tailored explanation. A prompt template combines instructions with runtime context. That context can include Salesforce record data and, depending on the configuration, resources from Flow, Apex, related lists, or Data 360. Reusable templates help teams keep instructions consistent and let authorized users execute a reviewed prompt within Salesforce workflows.
Start by separating a generation task from a deterministic task. Prompt Builder fits when the output needs natural language, context-sensitive wording, synthesis of several inputs, or a draft that a person can review. For example, a service representative may need a concise case summary using case details and recent interactions. A sales user may need a first draft that reflects an opportunity's products and stage. In both cases, the template provides a stable instruction pattern while record-specific values change at run time. A formula or ordinary Flow is usually a better fit when the requirement is exact arithmetic, a fixed decision rule, or a repeatable update whose result must be identical every time. A prompt can interpret context and draft language, but an LLM can produce varied output. If an action must always enforce a limit, calculate a contractual amount, or update a sensitive field under strict criteria, keep that rule in deterministic logic and use a prompt only for the language or reasoning task it can safely support.
Prompt Builder is also different from an agent. A prompt template produces content for a defined task. An Agentforce agent can interpret a request, select among topics and actions, and continue through a conversation or task. An agent may call a prompt template as one part of a larger workflow, but the two are not interchangeable. When the business need is autonomous task routing or a conversational assistant that chooses actions, assess Agentforce as well as Prompt Builder.
Define the task before building
Write down the user, trigger, expected output, data needed, review step, and risk if the output is wrong. Then check whether the template's supported type can receive the necessary inputs and return the needed result. Consider whether the data is current, whether users have access to it, and whether generated content needs human approval. A prompt that lacks relevant context may sound plausible while being inaccurate. A prompt that receives excess context can expose data unnecessarily or make the instruction harder to follow.
In a scenario question, underline the requested output. If it is a generated summary or draft using Salesforce context, Prompt Builder may fit. If it is a fixed calculation or exact record update, look for deterministic platform automation.
Prompt Builder supports several template types and integration paths. The exam asks you to choose the appropriate capability, not assume every generative AI requirement should use a prompt template. Extra concept: Human review is a design choice, not a guarantee of correctness. For consequential actions, keep approval or validation steps in the process.
Official reference: Salesforce Help, Prompt Builder - https://help.salesforce.com/s/articleView?id=ai.prompt_builder_about.htm&language=en_US&type=5
Key takeaways
- Prompt Builder creates reusable prompts for context-aware generated content.
- Use deterministic logic for exact calculations and rules that must not vary.
- An agent may use a prompt, but a prompt template is not itself an autonomous agent.
