Power Automate AI Prompts: Create Custom Prompt With Text with GPT

In this “Power Automate AI Prompts” article, I will show you how to create a custom prompt using the “Create Text With GPT using a Prompt” features of AI Hub prompts. First, I will create a custom prompt, and then I will show you how to consume this custom prompt from your Power Automate flow. Before getting into the article demo, let’s understand what prompts and custom prompts are in Power Automate AI Hub.

Prompts in Power Automate AI Hub?

First, let us understand about the prompts in Power Automate.

What is “prompts” in Power Automate AI Hub?

In simple terms, “prompts” in the Power Automate AI Hub are like instructions or questions that help you interact with an AI model. When you’re working with AI in Power Automate, prompts guide the AI on what information you’re looking for or what task you want it to perform.

For example, if you’re using an AI model to analyze customer feedback, a prompt could be something like, “What is the overall sentiment of this review?” This prompt tells the AI to focus on determining whether the feedback is positive, negative, or neutral.

In short, prompts make it easier to communicate with AI models by providing clear, specific instructions, helping the AI understand what you need.

I would recommend you read my previous article where I have explained AI prompts, AI models, and document automation in detail: AI Hub in Power Automate: Power of Prompts, AI Models and Document Automation

What is a custom prompt in Power Automate AI Hub?

A custom prompt in the Power Automate AI Hub is a tailored instruction or query that you create specifically to interact with an AI model according to your unique needs. Unlike standard or pre-defined prompts, custom prompts are designed to address particular tasks or extract specific types of information relevant to your business or workflow.

When should we create custom prompt?

You should create a custom prompt in Power Automate AI Hub when you need:

Specific Instructions: Your use case requires detailed or specialized instructions that aren’t covered by default prompts.

Tailored Responses: You want the AI to provide responses or actions based on your unique data or requirements.

Enhanced Accuracy: You need to refine the AI’s output to be more relevant to your specific scenario or industry.

Use Cases for Custom Prompts

1. Customer Support Automation:
   • Scenario: You want to automate responses to customer inquiries based on specific types of questions.
   • Custom Prompt Example: “Identify the customer’s issue from this email and suggest a response based on our troubleshooting guide.”
2. Invoice Processing:
   • Scenario: You need to extract and categorize specific details from invoices.
   • Custom Prompt Example: “Extract the invoice number, date, and total amount from this scanned invoice.”
3. Feedback Analysis:
   • Scenario: You want to analyze customer feedback to understand common themes or sentiments.
   • Custom Prompt Example: “Analyze this customer feedback for mentions of product quality and summarize the overall sentiment.”
4. Content Generation:
   • Scenario: You need to generate marketing copy or product descriptions based on specific inputs.
   • Custom Prompt Example: “Create a product description for this item using the following features and benefits.”
5. Document Review:
   • Scenario: You want to review and categorize documents based on their content.
   • Custom Prompt Example: “Classify this document as a contract, report, or memo and extract key terms.”
6. Data Entry Automation:
   • Scenario: You need to automate the entry of data into a CRM system from different sources.
   • Custom Prompt Example: “Extract customer contact information from this form and enter it into our CRM system.”

Custom prompts help you fine-tune how the AI interacts with your data, making the automation more precise and aligned with your specific business processes.

Demo: Create a custom prompt and consume it from Power Automate flow

Use Case Scenario: Here I will automate the candidate hiring process in an organisation where the company posts about new requirement details and job seekers will send their resumes to the human resources or hiring department. Once the resume comes in to the company’s internal system, either via email, the SharePoint site, or any other application, the Generative AI will process the candidate resume and will decide whether to shortlist the candidate for further discussion.

The Generative AI will handle it automatically; hiring people don’t need to spend time on reading and short listing the candidate resume. For this Gen AI automation, I will create a custom prompt, and then I will use that prompt in my Outlook or SharePoint list trigger event.

Pre-requisites: Create List in SharePoint Online

In this demo, I will use SharePoint Online list, so I need to create a list with the name something like Candidate Hiring with the two new multiline text type columns, Job Description Text and Candidate Resume Text, and the Title column with the requirement title name.

My SharePoint list is ready with the three columns: title, job description text,  and candidate resume text.

Create a new candidate hiring custom prompt in Power Automate AI Hub

Login to your Power Automate portal.

From the AI Hub menu, click on the “Prompts” tile link. If you don’t see the “AI Hub,”  you will get it from the “More” option.

Then, from the “Prompts” selection page, click on this tile link “Create Text with GPT using a prompt”.

The prompt creation page opens.

Below is the prompt I am going to create:

Prompt name:

New Candidate Hiring Process Custom Prompt

Prompt Description:

Imagine you’re an HR recruiter, and you need to evaluate the resume in the “varCandidateResumeText” variable against the job description in the “varNewJobDescription” variable. Score the candidate from 1 to 5, where 5 represents a perfect match and 1 indicates a lack of fit. Provide a bullet-point list of the qualifications and requirements the candidate meets and another list of what they don’t meet. Additionally, create two more lists: one with the candidate’s top strengths and another with their weaknesses.

Include a summary at the beginning, stating the score and a recommendation on whether to hire the candidate. Generate the output in HTML code format.

Configure the following as shown in the below screenshot:

• Prompt name: Name your prompt.
• Prompt: Enter your prompt text.
• Create your variable: Click on the insert icon and create your variable by clicking on the add input link. Here you don’t need to pass the value; you just need to create the variable as you use in your prompt; the value will be passing dynamically.

Click on the “Test prompt” button to see how it looks and works.

Click on the “Save custom prompt” button.

My custom prompt has been saved.

I can see my custom prompt in “My prompts” library.

By now I have successfully created my new candidate hiring process prompt. Next, I want to call this custom prompt from my Power Automate flow; let’s see how we can call.

Consume or Call Custom Prompt from Power Automate Flow

Create an automated cloud flow with the trigger “When an item is created in SharePoint”.

Configure your site URL and list name in the “When an item is created” trigger. Then, add a create text with GPT using prompt action.

Select your custom prompt from the list as shown below:

Configure the following parameters for the “New Candidate Hiring Process Custom Prompt”:

• Candidate Resume Text: Select the “Candidate Resume Text” column from your candidate hiring list.
• New Job Description: Select the “Job Description Text” column from your candidate hiring list.

Next, add start and wait for an approval action.

Configure the start and wait for an approval as below; the assignee should be (assigned to) whom you want to send this to for review.

Next, add a condition control and complete the rest of your flow as shown below:

Configurations:

Configure the following parameters:

• Outcome in Condition Control: This is the outcome from Start and wait for an approval action.
• To in Send Email (V2): HR email or whom you want to send the status of a shortlisted candidate for interview.
• Title in Send Email (V2): Title of the email; this dynamic title was selected from the candidate hiring list.
• Email body: This is the output of Create text with GPT using a prompt action.

Below is my complete flow:

Let’s now test this custom prompt flow.

Add a sample. Job title, job description, and candidate resume text in the candidate hiring list.

Added sample job and resume details to the candidate hiring list:

Wait for a few seconds to see your flow trigger status:

The flow has been triggered and running. The approver will get the task in three places. Outlook email, Approvals module in Power Automate flow, and Microsoft Teams, the approver can approve or reject it from any of the tools.

I will approve it from the flow approvals module.

Review and click on the confirm button before your approval.

Now, the approval is done.

Next, the concerned person in HR or the recruitment team will receive an email about the job requirement candidate resume shortlisting status. Here, as I have configured my email, I will receive an email about that; let’s check it.

And, I have received the below candidate short listing status Generative AI prompt-based generated email:

The demo is completed now. Thank you!

Summary: Power Automate AI Prompts – Create Custom Prompt With Text with GPT Prompt

Thus, in this article, we have learnt how to create a custom prompt with text with the GPT Prompt Power Automate action and how to call a custom prompt from a Power Automate flow. In the demo, we have learnt how we can automate the candidate hiring process for a new job requirement in an organisation. This prompt uses the Generative AI model in Power Automate.

About Post Author

Microsoft Techy

As a SharePoint and Power Platform Maven, my greatest joy comes from sharing my expertise with colleagues, friends, and the tech community, helping them navigate the ever-evolving world of technology and guiding them through the dynamic landscape of modern technology.

See author's posts