Create a prompt

Custom prompts give makers the freedom to instruct the generative AI model to behave in a certain way or to perform a specific task. By carefully crafting a prompt, you can generate responses that suit your specific business needs. This transforms the model into a flexible tool to accomplish various tasks.

Supported languages

The following list of supported languages for prompts in AI Builder is subject to be updated:

Chinese (Simplified), Czech (Czech Republic), Danish (Denmark), Dutch (Netherlands), English (United States), Finnish (Finland), French (France), German (Germany), Greek (Greece), Italian (Italy), Japanese (Japan), Korean (Korea), Polish (Poland), Portuguese (Brazil), Russian (Russia), Spanish (Spain), Swedish (Sweden), Thai (Thailand), Turkish (Türkiye)

Use specific text for more relevant responses

The goal of prompt engineering is to create an instruction that's as specific as possible to get a more relevant response from the model. Your prompts should be specific to a topic and convey your intent.

A prompt might include the following information:

• The topic
• Keywords or phrases that are associated with the topic
• The tone of the response
• The target audience

If the generated text is too long or contains irrelevant information, adjust the prompt. A good prompt has the following characteristics:

• Clear and concise: It's written in clear and concise language that's easy to understand.
• Specific: It's specific enough to guide the GPT model in the right direction.
• Contextual: It provides enough context for the GPT model to generate meaningful output.
• Relevant: It's relevant to the task and provides the GPT model with enough information to generate meaningful output.

Parts of a prompt

There are generally two parts to a prompt: the instruction and the context.

• The instruction is the first part of the prompt. It should provide clear directions on what the GPT model should do, for example, Summarize this email in three bullets.
• The context is the second part of the prompt. It should provide the information the model needs to generate an appropriate response, for example, The email contains customer feedback from the past week.

Example

A custom prompt can guide the model to answer a question, complete text, translate languages, summarize a document, and identify tasks, to-dos, and action items in text. The complexity of a custom prompt can range from a single sentence to something more intricate, depending on the task.

Download: AI Builder prompt engineering guide

Download the guide here: AI Builder prompt engineering guide (10 pages, 10X13 in.)

Download and print the AI Builder prompt engineering guide (10 pages, 10X13 inch size) to keep it handy and get help creating prompts.

Create a prompt

Prompt builder provides the flexibility to create your own custom prompt by defining input variables that enable incorporating dynamic runtime content within a prompt. It gives you the ability to validate the prompt, ensuring optimal prompt performance and correctness of the response before integration into your business solutions.

1. Sign in to Power Apps, Power Automate, or Copilot Studio.

2. For Power Apps and Power Automate, select AI hub > Prompts > Build your own prompt.

3. For Copilot Studio, there are multiple ways to create a prompt:

   1. Select Agents > pick or create an agent > Actions > Add an action > New action > New prompt.
   2. Select Agents > pick or create an agent > Topics. Pick or create a topic, add a node, and then select Add an action > New prompt.
   3. Select Library > Add new > Prompt. Prompts created in the library can be used in any agent action or topic.

   

Configure and test your prompt

1. Enter a custom name for your prompt by selecting on the generated name at the top left of the prompt.

2. Write your prompt or choose a prompt template.

   Tip

   To access a larger set of prompt templates, go to the Prompt directory Sample Solution Gallery.

3. Include an input object by typing / or selecting Add content, and then select Text or Image or document in the Input section. Inputs allow you to pass dynamic values to the prompt at runtime.

   

   In this view, you can also select a knowledge object in the Knowledge section. A knowledge represents a set of data retrieved through a data source connection, like Dataverse. This allows your prompt to generate answers that are related to your business data.

4. Type a sample value for each input.

   

5. Select Test to run the prompt and review its response.

6. When you're satisfied, select Save.

Use your prompt in Power Apps or Power Automate

The app you plan to use with your prompt determines your next step.

• Use in Power Apps

  • Empowers makers to incorporate existing prompts into their apps.
  • Use a prompt in Power Apps

• Use in Power Automate

  • Empowers makers to incorporate existing prompts into their flows.
  • Use a prompt in Power Automate

Check out this short video to learn how to include your prompt in a flow.

Related information

Human review for automation with a prompt