Prompts
Imagine you just hired a highly talented assistant chef at your bakery. If you simply tell them to "bake a cake," you might not get what you want. But if you specify the flavor, size, occasion, and decoration guidelines, they will nail it every time.
Generative AI works exactly the same way. The instructions you give to a Large Language Model (LLM) are called prompts. The more detailed and effective your prompt, the better the AI's response will be.
The Recipe for a Perfect Prompt
If you are a bakery owner wanting to write a promotional blog, don't just ask the AI to "write a blog post." Instead, break your prompt down into these core characteristics:
- Participant&Setting: Define who you are. (e.g., "I am a bakery owner who loves blogging about cake decoration.")
- Goal: State your clear objective. (e.g., "I want to teach my customers how to design decorative cakes.")
- Relationship: Identify your target audience. (e.g., "This is for our regular customers.")
- Data: Provide the core context. (e.g., "We are launching a new cake collection.")
- Instructions: Set strict boundaries. (e.g., "Write in simple English that a 6-year-old can understand, and keep it under 600 words.")
Once you craft the perfect prompt, you can save it as a prompttemplate with placeholders. When you launch a new cake in the future, just swap out the data placeholder, and your AI is ready to go!
Bringing Prompt Templates to Salesforce
This exact strategy is why Salesforce introduced PromptBuilder. Instead of starting from scratch, Salesforce provides predefined templates tailored for business needs:
- SalesEmailTemplates: To quickly draft effective customer outreach.
- FieldGenerationTemplates: To automatically populate custom data fields.
- RecordSummaryTemplates: To instantly condense complex CRM records.
- FlexTemplates: For entirely custom business requirements.
By securely grounding these templates with your CRM data points—like MergeFields,Flows,Apex,andDataCloud—Salesforce delivers incredibly accurate, personalized responses.
