In other parts of this series, we cover Data Cloud and how it harmonizes customer data across multiple systems, the Einstein Trust Layer and how it keeps your business data secure, and Agentforce which uses those features along with our next focus – Prompt Builder.
In a nutshell, Prompt Builder simplifies how businesses work with generative AI. Powered by Salesforce’s trust layer, Prompt Builder allows businesses to create customized, reusable prompts for large language models (LLMs). Whether it’s crafting personalized emails or generating customer summaries, Prompt Builder transforms how businesses interact with AI, enabling more dynamic, data-driven outputs that drive efficiency and productivity.
Does Prompt Builder need Data Cloud? Not necessarily, but as we’ll explore, the more data you have, the more powerful your prompts become. The context that data provides helps tailor AI-generated outputs to your specific customers and use cases. In this article, we’ll walk through how Prompt Builder works and why it’s transforming everyday tasks for businesses.
Introducing Prompt Builder: A Smarter Way to Work with AI
At its core, Prompt Builder helps you communicate with large language models (LLMs), the underlying technology behind generative AI. The LLM takes in a prompt—a combination of instructions, guidance, and relevant data—and generates a response.
Specificity is key to creating effective prompts. A well-constructed prompt includes contextual information, clear instructions, and constraints, all of which guide the AI to produce relevant, accurate, and efficient outputs.
For example, if you ask an AI to “draft a follow up email to a customer,” it might return a generic response. You’d need details to ground the response in your use-case. Things like:
- Contextual information about the sender and recipient
- Your business and the products you offer
- The goal or outcome you hope to achieve with this email; is it a cold email or an upsell opportunity
- The relationship you have with the customer such as if they’re new or existing and what your previous interactions with them have been.
Beyond that contextual information, your instructions would need to be clear and detailed as well. For example, instead of “write a follow up email to a customer,” detailed instructions could be, “write a follow up email to a customer that highlights how our products would address their specific needs and pain points, the benefits of a long-term relationship with our business, and includes a call to action to schedule a meeting with a sales representative.”
Finally, you’d want to add constraints to further hone the output such as style and tone guidelines and guardrails to prevent it from veering from the given instructions or generating answers that aren’t explicitly based on the data or information you’ve provided.
It takes effort to construct this individual prompt effectively, but once you’ve achieved that, it’s difficult to apply it at scale across multiple scenarios and customers. That’s where the power of Prompt Builder comes in. Instead of manually creating these detailed prompts each time, Prompt Builder allows you to turn these instructions and constraints into reusable templates across thousands of customer interactions, and dynamically enrich your prompts with data specific to each customer.
This dynamic grounding makes every interaction more precise and tailored to the individual, without requiring you to start from scratch for each customer. It’s like having a “smart template” that automatically adjusts based on the unique attributes of each situation.
Dynamic Grounding and Secure Data Retrieval: Contextual AI with Trust
Dynamic Grounding and Secure Data Retrieval are core elements of Prompt Builder built on the Einstein Trust Layer. These are just some elements of the trust layer, which we cover more thoroughly in the Einstein Trust Layer blog post of this series, but it’s what makes prompts adaptive while respecting your data sensitivity and user permissions.
Dynamic Grounding makes it easy to fill prompts with real-time, context-specific data from your Salesforce CRM. This feature transforms a generic prompt into a personalized customer interaction, ensuring that each customer’s unique details are always reflected in your communications. Think of it as using merge tags—placeholders that dynamically change based on the scenario.
Take the example of a personalized email to a customer. With dynamic grounding, placeholders for the customer’s name, recent interactions, and location are automatically filled in, based on the data in your Salesforce CRM. This means that when interacting with John Doe, the AI will use his name, address, and customer history; when interacting with Jane Smith, the prompt dynamically adjusts to her information.
Even more powerful, this grounding updates as interactions evolve. Let’s say you’re using AI in a customer support scenario. The placeholder for “conversation history” would refresh each time the customer sends a new message, ensuring that responses are relevant to the current state of the conversation.
At the same time, Secure Data Retrieval ensures that this dynamic data is retrieved and used securely. Operating within Salesforce’s trust layer, data access respects all existing user permissions and organizational security settings. This means prompts grounded in sensitive business data only retrieve what authorized users are permitted to see. In this way, Prompt Builder boosts productivity while maintaining a high level of data security.
Different Prompt Types in Salesforce
Salesforce’s Prompt Builder offers several prompt template types to fit different business needs:
- Sales Email Templates: These allow you to draft personalized emails automatically by grounding the content in customer data from your CRM. Whether it’s a welcome email or a follow-up on a recent purchase, these templates ensure that every communication is tailored and relevant.
- Field Generation Templates: This enables single-click actions to generate descriptions, summaries, or other field values. For example, an AI-generated product description can be produced in seconds, based on input data from your CRM.
- Record Summary Templates: These are used to create custom summaries of records, drawing on field data, activity history, and related records. This is particularly useful for providing agents or sales teams with quick overviews of customer accounts or cases, ensuring they have the most relevant information at their fingertips.
- Flex Templates: Flex templates are versatile, capable of grounding prompts on multiple objects, and can be applied to use-cases not covered by other prompt template types. This means they can be called as invocable actions within any business process, making them a flexible tool for complex workflows.
Final Thoughts
Salesforce’s Prompt Builder offers a clear advantage by streamlining routine tasks that can bog down teams. Whether it’s writing emails or generating customer summaries, tasks that typically require manual effort are now automated and grounded in real-time data. For instance, need to draft a sales email? With a pre-configured Sales Email Template, Prompt Builder can do it in seconds, pulling in relevant customer details like purchase history or recent interactions to deliver a personalized, timely response.
In summary, Prompt Builder offers a powerful way to harness AI for real-world business tasks, helping teams streamline workflows and increase productivity. By automating routine tasks like writing emails and summarizing records, Prompt Builder allows businesses to focus on what truly matters—building better customer relationships. If you’re ready to explore the full potential of AI within Salesforce, our team of experts at Port & Starboard is here to help!
