Published at: 2026-09-17
Prompt template overview
Large language models (LLMs) can help enterprises and teams complete various natural language tasks. For example, they can generate personalized emails to send to customers, or analyze account feedback to extract key insights. However, an LLM cannot produce great content on its own. It needs some guidance from us in the form of prompts.
A ShareCRM prompt template is a reusable prompt. The template contains placeholder variables for filling in specific details such as customers and products. Once you fill these placeholder variables with real, relevant CRM data, the prompt template becomes a truly personalized prompt.
Elements of a prompt
- Instruction: The specific task or instruction you want the model to perform.
- Context: External information or additional context that guides the model to respond better.
- Input data: The content or question the user provides.
- Output instruction: Specifies the type or format of the output.
General tips for designing prompts
Start simple
You can start with a simple prompt and gradually add more elements and context as you want better results. Therefore, iterating on your prompt throughout this process is essential. As you read this guide, you will see many examples where specificity, conciseness, and brevity usually lead to better results.
When you have a large task that involves many different subtasks, try breaking the task into simpler subtasks and building up gradually as the results improve. This avoids adding too much complexity at the beginning of the prompt design process.
Instructions
You can use commands to instruct the model to perform various simple tasks, such as “write,” “classify,” “summarize,” “translate,” and “sort,” to design effective prompts for a variety of simple tasks.
Remember that you also need to experiment a lot to find the most effective approach. Try different instructions with different keywords, contexts, and data to see what works best for your specific use case and task. In general, the more specific and relevant the context is to the task, the better the results.
Specificity
Be very specific about the instructions and tasks you want the model to perform. The more descriptive and detailed the prompt is, the better the results. This is especially important when you have requirements for the generated results or style. More importantly, have a well-formatted and descriptive prompt. In fact, providing examples in the prompt is very effective for getting the desired output in a specific format.
When designing prompts, also pay attention to the prompt length, because prompts have a length limit. Consider how specific and detailed you need to be. Including too many unnecessary details is not necessarily a good approach. The details should be relevant and help complete the task at hand. This is something you need to experiment with a lot. We encourage extensive experimentation and iteration to optimize the prompt for your application.
Avoid ambiguity
Given the advice above about being detailed and improving format, it is easy to fall into the trap of trying to be too clever with your prompt, which may create ambiguous descriptions. In general, it is better to be specific and direct. The analogy here is much like effective communication - the more direct you are, the more effectively the message is conveyed.
What to do, not what not to do?
Another common tip when designing prompts is to avoid saying what not to do, and instead say what to do. Saying what to do is more specific and focuses on the details that help the model generate a good response.
Use delimiters
Delimiters can help separate different parts of the text so that the model can better understand them for different processing. Delimiters can be any special character sequence that usually does not appear together. Some examples are:
###, ===, >>>