AI Hub: Build Resumes Reviewer App via AI Prompts (AI Builder)
What is AI Prompts
There have been significant feature updates to both Microsoft 365 services and the Power Platform for a while. Most of them are related to AI-based updates, including co-pilot, which makes working easy and fun. This year, artificial intelligence-supported Power Apps applications began to be made available to almost everyone.
AI Prompts
t does not really matter whether you previously used AI Builder or not; you can easily handle how to work on it. All you need to do is create a developer account with your work or school account. Get more information about Microsoft 365 developer sandbox subscription.
Get more information about Microsoft 365 developer sandbox subscription.
You can create 5 different types of AI based feature on Power Platform.
1- Create text with GPT using a prompt
2- Respond to a complaint
3- Classify Text
4- Extract information from text
5-Summarize Text
In this article we are going to have a look closer one of them:
1-Create text with GPT using a prompt
According to official Microsoft documentation description of Create text with GPT using a prompt:
This model runs on the Azure OpenAI Service and can be used for many tasks that involve creating text. The prebuilt prompts can use generative AI in several handy scenarios. You can also create a custom prompt for more custom behavior. When you’re done, you can use the model in an app or flow.
Talk is cheap let’s go to action
Before jump to Power Apps, first of all we need to create a prompt what we decide to do. In a nutshell we are going to create an application that reviewing the candidate’s applying of resumes as human resources aspect on canvas app.
What will the purpose of the canvas application?
The HR recruiter will upload the resumes to the app. They will automatically be turned into text, meanwhile, the HR person can write or copy and paste what they expect in terms of qualifications and requirements from the candidate. The remaining part of the process will be left in the hands of the AI prompt: According to the inputs that we provide, the pre-built model is going to evaluate the two texts and return texts about the strongest parts of the candidate, weaknesses, and additionally, the model will provide the text of the matched or not matched qualifications and requirements of the candidate that are expected by the HR person. Finally, it will evaluate the candidate’s requirements and give a rating from 1 to 5. As a result, it will suggest whether the candidate deserves to be hired or not, or if you like, you should give them a beginner crash course or something similar.
Explaining step-by-step Creating Prompt
It’s really important to see accuracy phrase and sentences that expectation of the user that creating the prompt. If you do not provide detailed information about what exactly see as returned value, the output will not relevant that your expectation or not satisfied you.
Prompt: Act as a HR recruiter, evaluate the given resume from HR perspective.
[VARIABLE]
and compare it against the position requirements:
[VARIABLE]
Give the candidate a ranking from 1–5 based on how closely they meet the job requirements and qualifications with being 5 perfect match and 1 being not a match. Provide short information of the candidate. Include a bullet point list of the qualifications and requirements that the candidate has which match our provided requirement and another bullet point list of the what the candidate does not have in our qualifications. Include two additional lists, one that lists the top strengths of the candidate and another that lists weaknesses. First list name should be Top Strength of the Candidate and second list name should be Weaknesses of the Candidate. Include a TLDR summary at the top of the score and recommendation to hire or not hire the candidate. Always use this sequence in the list: TLDR Summary, Ranking, Qualifications and Requirements Match, Qualifications and Requirements Not Match, Top Strengths of the Candidate, Weaknesses of the Candidate, Recommendation
Given prompt at above example has two variable. You can add any string wherever part of the text. Variable placement is important because it will affect the quality of the returns text. Second thing that required attention is be aware of the given text is enoughly detailed or not. As you can see of this phrase expectation was highlighted to the direction of the sequence of the expectation titles.
When you complete placement of the text and the variable that you need, then second thing that you doing is to test it.
All you need is just adding your resumes converted text to ResumeText textbox and requirements to PositionRequirements textbox.
AI response gave us some of the outputs through the expectation that we provided in the texts. In order to get exact output I provided as much as detailed information: TLDR Summary,Ranking, Qualifications and Requirements Match, Qualifications and Requirements Not Match, Top Strengths of the Candidate, Weaknesses of the Candidate.
If the test is successful for you you can save the custom prompt.
Create Power Apps Canvas App
We covered the AI part of the app, now time to show the output somewhere. In this article we will serve it on canvas application, but you can use the output on Power Automate or co-pilot powered bots or any Power Platform services.
You can add just some landing page components that giving information about the application. I prefered add header, main content which made of three sub components: welcome text, wizard gif which is reflecting the purpose of the application and a button due to navigate to resumes evaluate screen
You will navigate to first step of the resumes evaluation screen when you click on Get Started button
All you need are adding the fundamental components which provides inputs from the user. Two of them are text inputs which will provide the information about the candidates resumes and expectations which is typed by HR team person.
Remaining two components are very important to know how to achieve get text string from the uploaded pdf file. It’s important part of the provided resumes are going to parsed as string, first state of the uploaded pdf file is record which is each of the sentences equal to one record.
I collected every single sentences as one record in a collection with concat function, normally it should not a record, just a multi text.
Back to screen and click on Resume Wizard button.
And see the prompts that the model returned!
You can see the AI Prompts and Power Apps collaboration in action at below!
See you next article!
