Chapter 4 · Role-Specific Scenarios

4.1 HR: JD Generation, Resume Screening & Candidate Follow-ups

Whether you're a recruiter or an HRBP, the hiring flow is similar:

StepTraditional approachPain point
1. Write job descriptions (JDs)Copy from old descriptions, reformat for each channelMultilingual, multi-format adaptation takes time
2. Screen resumesOpen each one, score manuallyLow efficiency at volume, easy to miss candidates
3. Manage candidatesExcel tracking, easy to forget follow-upsCandidates go cold after a week without contact

This lesson uses AOS Work to draft job descriptions, organize resumes, and write results to one Base. Recruiters publish the roles and arrange interviews.

Preparation

TypeResourceWhat it does hereNotes
AppBaseHolds job descriptions, candidate rankings, and interview feedbackPrepare job-description and candidate tables and specify them in the task
Skill (optional)Job-description writingReuses writing and channel-format rulesUse a writing Skill you have added, or use the prompt below directly
MaterialsRole information and practice resumesInputs for drafting and screeningPrepare anonymized PDF/Word resumes, role requirements, and scoring criteria

For enabling Base and reading or updating records in a task, see the Base exercise. Set up the job-description and candidate fields with Base tables. For reading original resumes, see the Library file exercise.

Sub-scenario A: Generate Multilingual Job Descriptions

Pain Point

"One role needs to go on LinkedIn, Indeed, Glassdoor, and our website. Each channel has a different format. Plus it needs Italian and English versions. Every time I spend half an hour reformatting."

Steps

  1. Prepare the role info.
    Tell the AI the following elements:
ElementDescriptionExample
Role nameIn English and local languageSenior Merchandiser
LevelJunior / Mid / SeniorSenior
Core responsibilities3–5 itemsMerchandise planning, SKU analysis, seasonal buying plan
RequirementsHard requirements5+ years fashion industry, fluent Italian and English
Nice-to-havesPreferred conditionsLuxury brand experience, SAP knowledge
  1. Specify output requirements.

    • Languages: Italian + English
    • Channel formats: LinkedIn / Indeed / Glassdoor / Company website
    • Special requirements: Exclude discriminatory language, generate formal and social-media versions
  2. AI generates the JDs.
    The AI creates multiple versions based on your input.

  3. Review and write to the Base.
    After confirming, the AI writes the JDs to the table for easy management and publishing.

Sample Prompt

Generate Italian + English JDs for the following role:
- Role: Senior Merchandiser
- Level: Senior
- Core responsibilities: Merchandise planning, SKU four-quadrant analysis, seasonal buying plan development
- Requirements: 5+ years fashion industry, fluent Italian and English, proficient in data analysis tools
- Nice-to-haves: Luxury brand experience, SAP knowledge
Output requirements:
1. Adapt for LinkedIn / Indeed / Glassdoor / Company website — four channel formats
2. Exclude discriminatory language (do not restrict applicants by age, gender, race, religion, or marital status)
3. Generate formal and social-media short versions
4. After review, write all versions to my specified job-description table in Base, recording role, channel, language, and version type

Exercise: Draft, revise and save job descriptions

This exercise drafts job descriptions for a Senior Merchandiser. Create English and Italian copy for LinkedIn, Indeed, Glassdoor, and the company website. Each channel gets a formal version and a shorter social-media version: 2 languages × 4 channels × 2 styles = 16 versions.

Work in the demo workspace named Product Team and select Base in the composer before sending the request. LinkedIn, Indeed, and Glassdoor are real recruitment or career platforms. The exercise prepares copy for these channels without publishing it to them.

Next, revise the English LinkedIn social version to mention flexible working. Save all versions to the Job Descriptions table in Base, then query and display the saved records.

Expected Results

  • 2 languages × 4 channels (LinkedIn / Indeed / Glassdoor / Website) × 2 versions = 16 JDs

  • All written to the Base, organized by channel and language

  • Ready to copy and publish directly from the table

For multiple roles, save the generated job descriptions in one table, grouped by role, channel, and language.

Sub-scenario B: Resume Screening and Match Ranking

Pain Point

"200 resumes for one role. If I spend 5 minutes on each, that's about 17 hours of screening. And I still miss great candidates."

Steps

  1. Upload resumes.
    Upload a batch of resumes (PDF/Word) to the AOS Work conversation.

  2. AI auto-parses.
    The AI reads each resume and extracts key info:

FieldDescription
NameFull name
ExperienceTotal years of work
SkillsIndustry tools, professional skills, software
LanguagesLanguages and proficiency
Last employerMost recent company
Current roleMost recent job title
  1. Input job requirements, AI scores the match.
    Provide the role requirements and scoring weights. Ask the agent for a 0–100 match score, resume evidence, and items to verify, then rank by score.

  2. Results written to the Base.
    The match ranking table goes into the table. Hiring managers can view, filter, and sort. Original resume files (PDF/Word) can also be linked to the corresponding records.

  3. Natural language follow-up queries.
    You can filter candidates directly with natural language:

Example query

List all candidates who meet these criteria: 3+ years of luxury-industry experience Italian at B2 or above currently based in Milan

Sort by match score, highest first.

Sample Prompt

Parse the following resumes, extract key info (name, experience years, skills, languages, education, current role).
Match against the following job requirements:
- Role: Senior Merchandiser
- Requirements: 5+ years fashion industry, Italian C1 or above, English B2+, proficient in Excel and data analysis tools
First list the scoring dimensions and weights for my confirmation. Then generate a match ranking table (0–100 points) in the specified candidate table. Mark missing information for verification without filling it in. Include:
Name, Experience years, Current role, Last employer, Skills, Languages, Match score, Rank, Scoring evidence, Items to verify.
Link original resume files to the corresponding records.

Expected Results

  • Candidate records contain extracted fields, match scores, and scoring evidence; missing information is marked separately

  • Base supports filtering and sorting — hiring managers can view directly

  • You can query specific candidates using natural language

Advanced: Update Interview Feedback

After the interview, type your evaluation directly in AOS Work:

Sample Prompt

Candidate Marco De Santis — interview feedback: strong industry experience, deep understanding of merchandising, fluent English, Italian needs improvement.
Result: proceeding to the second round.
Recommendation: schedule a technical interview with the team lead.
Update the corresponding Marco De Santis record in the specified candidate table with this feedback and result.

The agent writes the feedback and result to the candidate record. Open Base to check the update.

Advanced: Find Candidates Who Need Follow-up

Record candidate status and last follow-up date in the table, then query who needs a follow-up:

Sample Prompt

Query the specified candidate table for records with status "Pending" and a last follow-up date at least 7 days ago. Return name, last follow-up date, and next action, sorted by date ascending.

Exercise: Screen resumes and update interview progress

Use three candidate resumes for this exercise. Select Library and Base in the composer, read the resumes, and first agree on the criteria and weights for experience, languages, and Excel skills.

Save the ranking, scoring evidence, and original resume links to the Candidates table in Base. Filter candidates, record that Marco De Santis proceeds to a second interview, and update his follow-up date to 2026-10-26. Using that date, query pending candidates whose last follow-up was at least seven days ago.

HR Scenario Chain

These two scenarios aren't isolated. String them together into a complete hiring workflow:

Scenario Chain

Scenario A (Generate JD) → Publish to channels → Receive resumes
        ↓
Scenario B (Screen resumes) → Match ranking → Schedule interviews
        ↓
Interview feedback written to table → Candidate status updated

Job descriptions, candidate rankings, and interview feedback are saved in the same Base so you can review each candidate's record.

Summary

The agent drafts job descriptions and organizes screening evidence; recruiters review results and follow up with candidates.