Key Takeaways

The points worth keeping

• The Recruitment Funnel with AI at Each Stage
• 1. Resume Screening with AI
• 2. Optimized Job Descriptions

1. Resume Screening with AI

The most time-consuming and repetitive process in HR. A popular job posting receives 150-300 resumes. Reading every one carefully is impossible. Skimming leads to mistakes.

The real problem: it's not just volume. It's bias. Studies show that recruiters spend an average of 7.4 seconds on each resume during the initial screening. In that time, unconscious biases (name, university, photo, address) carry more weight than actual competencies.

The AI solution: create objective scoring criteria and let AI apply them consistently to every resume.

Prompt: Resume Scoring System

You are a resume screening analyst for a [JOB TITLE] position at [COMPANY].

Evaluation criteria (score 0-10 each):

RELEVANT EXPERIENCE (weight 3x):
- Years of experience in the role or similar
- Previous companies in the same industry
- Relevant projects or deliverables mentioned

EDUCATION AND CERTIFICATIONS (weight 2x):
- Academic background aligned with the role
- Relevant certifications
- Complementary courses

TECHNICAL SKILLS (weight 2x):
- Proficiency in required tools: [LIST TOOLS]
- Technical skills mentioned vs. required
- Evidence of results (numbers, metrics)

BEHAVIORAL SKILLS (weight 1.5x):
- Evidence of [DESIRED SKILLS: leadership, teamwork, etc.]
- Career progression (growth or stagnation)
- Stability vs. excessive turnover

CULTURAL FIT (weight 1.5x):
- Demonstrated values compatible with [COMPANY VALUES]
- Relevant extracurricular activities or volunteering
- Stated career goals

For each resume, deliver:
1. Total score (0-100)
2. Top 3 strengths
3. Top 2 concerns or gaps
4. Recommendation: ADVANCE / TALENT POOL / DO NOT ADVANCE
5. Customized interview question suggestion for this candidate

IMPORTANT: Ignore name, gender, age, photo, university (unless the
explicit criterion requires specific education). Evaluate only competencies
and demonstrated results.

Resume:
[PASTE RESUME HERE]

Practical Screening Flow

STAGE 1: Define criteria (human — 20 min)
→ Fill in the prompt above with the actual role requirements

STAGE 2: Initial screening (AI — 5 min per batch of 10)
→ Paste resumes in batches of 10
→ AI classifies: Advance / Talent Pool / Do Not Advance

STAGE 3: Human review (recruiter — 30 min)
→ Review ONLY the "Advance" candidates (typically 15-20% of total)
→ Confirm or reclassify
→ Check that AI didn't discard relevant atypical profiles

STAGE 4: Final shortlist (human — 15 min)
→ Select 5-8 candidates for interviews
→ Use the personalized questions suggested by AI

Typical result: screening that took 2 days drops to 3 hours. And with more consistency.

2. Optimized Job Descriptions

A poorly written job posting attracts the wrong candidate. It's the most expensive recruiting mistake because it contaminates the entire pipeline.

Common job posting problems:

• Inflated requirements ("5 years minimum experience" for an entry-level role)
• Language that pushes away diverse candidates (masculinized terms, exclusive jargon)
• Focus on requirements instead of value proposition
• Generic benefits that don't differentiate
• Too long or too short

Prompt: Job posting that attracts the right candidate

Create a job description for [JOB TITLE] at [COMPANY NAME], in the [INDUSTRY] sector.

Context:
- Company size: [X employees]
- Work model: [remote / hybrid / on-site in CITY]
- Salary range: [R$ X.XXX to R$ X.XXX]
- Reports to: [manager's job title]
- Team: [size and composition]

The description should have these sections, in this order:

1. HOOK (2-3 lines): Why should someone want this job?
   Don't start with "We're looking for." Start with the impact the person will have.

2. ABOUT THE COMPANY (3-4 lines): What we do, why we exist, culture in 1 sentence.

3. WHAT YOU'LL DO (5-7 bullets): Real responsibilities, not generic ones.
   Use action verbs. Include expected impact.

4. WHAT WE'RE LOOKING FOR (divided into 2):
   - ESSENTIAL (5 requirements max): What really eliminates if missing
   - NICE-TO-HAVE (3 items): Points that add up

5. WHAT WE OFFER (5-7 bullets): REAL and SPECIFIC benefits.
   Don't write "competitive salary." Write the range.
   Don't write "dynamic environment." Describe what that means.

6. HIRING PROCESS (3-4 stages): Transparency about the phases.

7. DIVERSITY (2 lines): Genuine statement, not generic.

Rules:
- Gender-neutral language
- Tone: professional but human (not cold corporate)
- Maximum 500 words total
- No unnecessary jargon
- No inflated "minimum requirements"

Tip: Run the description through AI a second time asking it to identify terms that may push away diverse candidates. Words like "aggressive," "ninja," "guru," "rockstar" push more people away than they attract.

3. Structured Interview Guide

An interview without a guide is a conversation. A conversation doesn't assess competence—it assesses likability. And likability introduces bias.

Structured interviews predict performance 2x better than unstructured interviews. AI creates consistent guides in minutes.

Prompt: Interview guide by level

Create a structured interview guide for [JOB TITLE], level [JUNIOR / MID / SENIOR / LEADERSHIP].

Competencies to assess:
1. [TECHNICAL COMPETENCY 1]
2. [TECHNICAL COMPETENCY 2]
3. [BEHAVIORAL COMPETENCY 1]
4. [BEHAVIORAL COMPETENCY 2]
5. [CULTURAL FIT]

For each competency, create:

MAIN QUESTION (behavioral — STAR method):
- A question that requires a real past example
- Format: "Tell me about a time when you..."

FOLLOW-UP QUESTION:
- To use if the answer is superficial
- Format: "How did you decide...?" / "What would you do differently?"

SITUATIONAL QUESTION (hypothetical):
- A realistic scenario the person would face in this role
- Format: "Imagine you [scenario]. How would you approach it?"

EVALUATION CRITERIA (scorecard 1-5):
1 = Did not demonstrate the competency
2 = Demonstrated partially, with significant gaps
3 = Demonstrated adequately for the level
4 = Demonstrated above expectations for the level
5 = Demonstrated exceptional mastery with clear impact

POSITIVE SIGNALS (what to listen for):
- 3 indicators of a strong answer

RED FLAGS (warning signs):
- 3 indicators of a weak answer

Adjust depth according to level:
- Junior: focus on potential, curiosity, fast learning
- Mid: focus on autonomy, problem-solving, delivery
- Senior: focus on technical leadership, mentorship, strategic decision-making
- Leadership: focus on people management, vision, team results

Also include:
- 2 minutes opening (how to welcome the candidate)
- 5 minutes for candidate questions at the end
- Estimated total time: [45 / 60 / 90 minutes according to level]

Competency table by level

Use this table to calibrate questions and evaluation criteria by level.

4. Personalized onboarding

Generic onboarding wastes the first 90 days. The new employee receives a 40-page PDF, an office tour, and a "feel free to ask if you have questions." Result: 3-6 months to reach full productivity.

With AI, you create a customized path based on role, seniority, and context.

Prompt: 90-day onboarding plan

Create a 90-day onboarding plan for a new [JOB TITLE] at [LEVEL]
at [COMPANY NAME], in the [INDUSTRY] sector.

Context:
- Team: [size and composition]
- Direct manager: [job title]
- Tools used: [list main ones]
- Current team projects: [summarize]
- Culture: [describe in 2 sentences]

Plan structure:

WEEK 1: ONBOARDING
- Day 1: Hour-by-hour schedule
- Day 2-3: Meetings with whom (and objective of each meeting)
- Day 4-5: First symbolic deliverable (something simple but real)
- Assigned buddy/mentor: yes, with agenda for first conversation

WEEKS 2-4: IMMERSION
- Required training (technical + culture)
- Shadowing: with whom and for how many hours
- First real deliverables (increasing complexity)
- Weekly check-in with manager: suggested agenda

MONTH 2: PROGRESSIVE AUTONOMY
- Own projects with reduced supervision
- Participation in team rituals
- Informal 360 feedback (how to gather)
- Competency milestones: what should be mastered by end of month

MONTH 3: FULL PRODUCTION
- Independent deliverables expected
- Formal probation period evaluation
- Development plan for the next 6 months
- Onboarding retrospective: what worked, what to improve

For each week, include:
- Main objective
- Specific activities
- Responsible person
- Success criteria (how to know if it worked)
- Potential friction point and how to mitigate

Practical tip: generate the plan with AI, have the direct manager review it, and personalize it with team information. The new employee receives something that feels tailor-made — because it was.

5. Climate and engagement survey

The annual engagement survey is dead. By the time you get results, analyze them, and take action, 4-6 months have passed. Real engagement issues need weekly or biweekly pulses.

How it works with AI

1. Collection: 3-5 short questions per week (Google Forms, Typeform, or HR tool)
2. Analysis: AI processes open responses, identifies themes and sentiments
3. Alert: automatic flag when a negative theme appears 3+ times
4. Action: AI suggests interventions based on detected issue

Prompt: climate survey response analysis

Analyze the responses below from a weekly organizational climate pulse survey.

Company: [NAME], [SIZE] employees, [INDUSTRY] sector
Period: week of [DATE]
Total respondents: [N] of [TOTAL] (response rate: [X]%)

Questions and compiled responses:
[PASTE RESPONSES HERE — can be anonymized]

Deliver:

1. OVERALL SENTIMENT SCORE: Positive / Neutral / Concerning / Critical

2. TOP 3 POSITIVE THEMES:
   - What's working well (with anonymized quotes)
   - Recommendation: how to maintain or amplify

3. TOP 3 AREAS OF CONCERN:
   - What's causing dissatisfaction (with anonymized quotes)
   - Urgency level: low / medium / high
   - Suggested immediate action for each

4. COMPARISON (if historical data exists):
   - Improved, stable, or declined? On which themes?

5. ALERTS:
   - Signs of burnout or disengagement
   - Areas or teams needing special attention
   - Turnover risk based on comments

6. SUGGESTED ACTION PLAN:
   - 1 recognition action (this week)
   - 1 corrective action (within 15 days)
   - 1 structural action (within 30 days)

Format: short bullets, actionable language. No "considering possibilities" language.
Tell them what to DO.

Engagement metrics to track

6. Training and development with AI

The biggest waste in T&D: training everyone the same way. The intern sits through the same lecture as the manager. The developer who masters Python watches the basic Python course because "it's mandatory for everyone."

With AI, you create personalized content based on skill gaps.

AI-driven T&D flow

STEP 1: DIAGNOSIS (AI + manager)
→ Map role competencies vs. employee competencies
→ Identify priority gaps

STEP 2: CONTENT CURATION (AI)
→ For each gap, suggest:
  - 1 short resource (article, 10-min video)
  - 1 medium resource (2-4 hour course)
  - 1 hands-on activity (project, exercise, job rotation)

STEP 3: PERSONALIZED LEARNING PATH (AI + HR)
→ Build logical learning sequence
→ Define deadlines and checkpoints
→ Assign mentor or buddy for each path

STEP 4: PROGRESS TRACKING (AI)
→ Quiz or self-assessment after each module
→ Manager feedback on practical application
→ Path adjustment based on progress

STEP 5: IMPACT ASSESSMENT (AI + HR)
→ Compare performance before and after
→ Calculate training ROI
→ Document for future cycles

Prompt: create internal training content

Create a micro-training (20-30 minutes) on [TOPIC] for [AREA/ROLE] employees.

Audience level: [BEGINNER / INTERMEDIATE / ADVANCED]
Format: structured text to present or read (no slides needed)

Structure:

1. OPENING (2 min):
   - Why this matters in their daily work (not theory, practice)
   - 1 data point or example showing the impact

2. CORE CONCEPT (5 min):
   - Explain the concept in plain language
   - Use an everyday analogy
   - Maximum 3 key points

3. REAL EXAMPLE (5 min):
   - Practical case (preferably from [SECTOR] industry)
   - What worked and why
   - What went wrong and what was learned

4. PRACTICAL EXERCISE (10 min):
   - Activity the participant does DURING the training
   - Must be applicable to their actual work
   - Tangible result at the end

5. APPLICATION CHECKLIST (3 min):
   - 5 actions the participant can take this week
   - Nothing vague. Everything specific and actionable.

6. RETENTION QUIZ (5 min):
   - 5 multiple choice questions
   - Answers with brief explanation

Tone: direct, engaging, no excessive teaching style.
Language: EN, professional informal.

7. Feedback and performance review

Annual performance reviews are a fiction. Nobody remembers what they did in March when the review happens in November. And the manager evaluating 12 people basically writes the same text for everyone.

AI transforms reviews into a continuous, evidence-based process.

Prompt: generate performance review

Generate a performance review for [EMPLOYEE NAME], role [JOB TITLE],
period [MONTH/YEAR to MONTH/YEAR].

Input data (collected throughout the period):

DELIVERABLES:
[List main deliverables, completed projects, and results]

GOALS AND RESULTS:
[Goals set at period start and % achievement]

FEEDBACK RECEIVED (from peers, manager, internal clients):
[Paste feedback collected throughout the period]

COMPETENCIES ASSESSED:
1. [Competency 1]: rating [1-5] from manager
2. [Competency 2]: rating [1-5] from manager
3. [Competency 3]: rating [1-5] from manager
4. [Competency 4]: rating [1-5] from manager
5. [Competency 5]: rating [1-5] from manager

Based on this data, generate:

1. PERFORMANCE SUMMARY (3-4 lines):
   - Overall performance in 1 sentence
   - Main achievement of the period
   - Main development area

2. STRENGTHS (3-5 bullets):
   - Based on concrete evidence (deliverables + feedback)
   - Each point with specific example

3. DEVELOPMENT AREAS (2-3 bullets):
   - Constructive, not punitive
   - With action suggestion for each area
   - Suggested timeline for improvement

4. INDIVIDUAL DEVELOPMENT PLAN (IDP):
   - 2-3 development actions for next period
   - Suggested resources (course, mentorship, project, job rotation)
   - Success criteria for each action

5. RECOMMENDATION:
   - Promotion / Merit / Maintain / Next period follow-up
   - Justification in 2 lines

Tone: fair, constructive, fact-based. No euphemisms that hide problems, but no aggression either. The goal is development, not punishment.

IMPORTANT: This is a DRAFT for the manager to review and personalize.
Never deliver an AI-generated review without human review.

Best practices for continuous feedback with AI

8. Offboarding and Exit Interview

The moment when the company can learn the most and most often ignores. Employees who leave tell the truth — if you ask properly.

Prompt: Exit Interview Script

Create an exit interview script to collect honest and actionable information.

Context:
- Termination type: [VOLUNTARY / INVOLUNTARY]
- Employee position: [JOB TITLE]
- Time at the company: [X years/months]
- Department: [DEPARTMENT]

The script should have 3 blocks:

BLOCK 1: OVERALL EXPERIENCE (10 min)
- What did you value most during your time here?
- If you could change 1 thing about the company, what would it be?
- How would you describe the culture to someone outside?

BLOCK 2: MANAGEMENT AND TEAM (10 min)
- How was your relationship with your direct manager?
- Did you feel you had opportunities for growth?
- What would make the team work better?

BLOCK 3: REASON AND FUTURE (10 min — only for voluntary termination)
- What mattered most in your decision to leave?
- What could the company have done to make you stay?
- Would you recommend this company to a friend? Why?

CLOSING:
- Is there anything you never had the opportunity to say?
- Is there anything you'd like us to know?

Interviewer rules:
- NEVER conduct the interview with the direct manager (conflict of interest)
- Ensure real confidentiality (not just a promise)
- Listen more, talk less (80/20 rule)
- Do not argue, defend, or justify. Just listen.
- Genuinely thank them at the end

Prompt: Turnover Pattern Analysis

Analyze the following exit interviews from the last [X] months:

[PASTE INTERVIEW SUMMARIES — ANONYMIZED]

Deliver:

1. EXIT PATTERNS:
   - Top 3 most cited reasons for leaving
   - Departments with the most departures
   - Average tenure before departure

2. WARNING SIGNS:
   - Systemic problems that appear repeatedly
   - Managers or areas with the most negative mentions
   - Topics that worsened over time

3. RECOMMENDED ACTIONS:
   - 3 short-term interventions (30 days)
   - 2 structural changes (90 days)
   - 1 cultural change (6 months)

4. TURNOVER COST ESTIMATE:
   - Average replacement cost by level (consider: recruitment,
     training, lost productivity, team impact)
   - How much the company would save by reducing turnover by 20%

Format: executive report, maximum 2 pages. Data language, not opinion.

LGPD in HR: What You Can and Cannot Do with Candidate and Employee Data

LGPD applies in full to HR. Candidates and employees are data subjects. Using AI does not change legal obligations — it just requires more care.

What You CAN Do

What You Should NOT Do

LGPD Checklist for HR

• Inform candidates that AI is used in the selection process
• Remove identifying data before processing with public AI
• Allow rejected candidates to request data deletion
• Document screening criteria (auditability)
• Use the business version of AI (that doesn't use data for training)
• Have a candidate data retention policy
• Ensure real anonymity in engagement surveys

Tools: Practical Comparison for HR

Free and Low-Cost Tools

Specialized HR Tools

Recommendation by Company Size

Bias and Fairness: How to Audit AI Decisions in HR

The biggest risk of AI in HR isn't inefficiency. It's amplifying biases that already exist.

If your historical hiring data has bias (and it probably does), AI will learn and replicate that bias at scale.

Most Common Biases in AI Recruitment

Quarterly Bias Audit (Checklist)

Every 3 months, run this analysis:

1. PIPELINE DATA BY DEMOGRAPHIC:
   - How many candidates from each group pass each stage?
   - Is there significant disparity between groups?

2. BLIND TEST:
   - Remove name, gender, age, and university from 20 resumes
   - Run AI screening again
   - Compare: did the results change?

3. HIRED VS. REJECTED ANALYSIS:
   - Are rejection criteria consistent?
   - Is there a demographic pattern among rejected candidates?

4. CANDIDATE FEEDBACK:
   - Do rejected candidates report fair treatment?
   - Are there recurring complaints from any group?

5. CALIBRATION:
   - Are evaluation criteria the same for everyone?
   - Do different interviewers give similar scores for similar answers?

Golden Rule: AI doesn't decide. AI informs. The final decision about people is ALWAYS human and documented.

Employee Experience: AI for Internal Communication and FAQ

Employees spend an average of 1.5 hours per day searching for internal information: "What's the health plan?", "How do I request time off?", "What's the remote work policy?".

An AI FAQ bot solves 80% of these questions instantly.

Prompt: Create an HR Knowledge Base

Create a comprehensive FAQ for employees at [COMPANY NAME] with [X] employees.

Required categories:

BENEFITS (10 questions):
- Health and dental insurance, meal vouchers, transportation vouchers
- Gym membership, childcare assistance, life insurance
- How to use, waiting periods, dependents

TIME OFF AND LEAVES (8 questions):
- How to request vacation, minimum notice required
- Maternity/paternity leave
- Medical leave, bereavement, wedding leave
- Birthday day off (if applicable)

WORK POLICIES (8 questions):
- Flexible hours, comp time
- Remote/hybrid work policy
- Dress code
- Use of company equipment

CAREER AND DEVELOPMENT (6 questions):
- How performance reviews work
- Criteria for promotion
- Course and certification subsidies
- Mentorship program

FINANCES AND PAYROLL (6 questions):
- Pay date, advances
- Pay stub (where to access)
- Profit sharing/bonus
- Expense reimbursement

For each question:
- Clear, direct answer (maximum 4 lines)
- Link or reference to complete document (if it exists)
- Who to contact if the answer doesn't help

Tone: Welcoming and clear. Like an experienced colleague is explaining.

HR Metrics: What to Measure and Why

If you don't measure it, you can't manage it. These are the metrics every AI-driven HR department should track.

Recruitment Metrics

People Management Metrics

Prompt: Monthly HR Dashboard

Based on the following HR data for [MONTH/YEAR]:

[INSERT DATA: headcount, hires, terminations, open positions,
average hiring time, engagement survey results, etc.]

Generate a 1-page executive report with:

1. EXECUTIVE SUMMARY (3 lines):
   - Overall HR status in 1 sentence
   - Main achievement of the month
   - Main concern

2. KEY METRICS (table):
   - Metric | Result | Target | Status (green/yellow/red)

3. HIGHLIGHTS (3 bullets)

4. AREAS OF CONCERN (3 bullets with suggested action)

5. OUTLOOK FOR NEXT MONTH:
   - Positions expected to open
   - Identified turnover risks
   - Planned T&D investments

Format: Visual, with numbers highlighted. Ready for executive presentation.

Case Study: Startup That Scaled from 10 to 50 People with AI in HR

Company: Technology startup in Florianopolis, SaaS B2B.

Initial situation: 10 people, no formal HR. The CEO handled recruiting, onboarding, and management "the hard way." Each hire took 45-60 days. Turnover was 30% in the first year.

What they did (without hiring HR for the first few months):

Months 1-3: Foundation

• CEO built a library of 15 HR prompts using ChatGPT Plus
• Standardized job descriptions (time: from 3 hours down to 20 minutes)
• Structured interview scripts by role (consistency across interviewers)
• Candidate evaluation scorecard (end of "gut feeling" decisions)

Months 4-6: Scale

• Adopted Gupy as ATS (R$400/month) with AI-assisted screening
• 30-day onboarding customized by role (generated with AI, refined by manager)
• Weekly pulse survey on engagement via Google Forms + analysis with Claude
• Hired 15 people in 3 months (time to hire dropped to 22 days)

Months 7-12: Professionalization

• Hired 1 HR person (who came in with processes already in place)
• Continuous feedback system: weekly check-ins with AI-generated agendas
• Semi-automated biannual performance reviews
• T&D customized by skill gaps
• Hired 25 more people. Turnover dropped to 12%.

Results over 12 months:

Key lesson: They didn't wait to have an HR department before building HR processes. They used AI to create the infrastructure and only hired a human when the volume justified it.

Common HR + AI Mistakes (and How to Avoid Them)

Mistake 1: Automating Without Humanizing

AI sorting resumes is efficient. AI sending a generic rejection email is inhumane. The rejected candidate could become a future client, partner, or referral source.

Solution: Use AI to personalize rejection communications. Three specific lines about the person's resume make all the difference.

Mistake 2: Blindly Trusting AI Scores

AI didn't see the nervous candidate who actually knows the material. It didn't notice the simple resume of someone who grew up without formal opportunities. Scores are filters, not decisions.

Solution: Manually review 10% of resumes rejected by AI for each open position. If you find good candidates that were discarded, adjust the criteria.

Mistake 3: Not Training Managers to Use AI

HR creates perfect interview scripts with AI. The manager ignores them and asks the same questions as always. The process becomes theater.

Solution: Two-hour practical training for managers: how to use the script, how to fill out the scorecard, and why it matters.

Mistake 4: Collecting Data Without Acting

A climate survey with no follow-up is worse than not surveying at all. Employees notice that "nobody does anything with the results" and stop responding.

Solution: For every survey, define 1 visible action within 15 days. Communicate what was done and why.

Mistake 5: Ignoring the Candidate Experience

If the hiring process is lengthy, confusing, and offers no feedback, the best candidates drop out. AI makes it possible to maintain frequent communication at no cost.

Solution: Automate status updates. "Your resume was received," "You've moved to the next stage," "The position has been filled." Simple and respectful.

Mistake 6: Using AI Without Transparency

Candidates and employees have the right to know that AI is involved in the process. Hiding it breeds distrust and can create legal issues.

Solution: Include in the job posting: "We use AI tools to assist with screening. The final decision is always human."

Implementation Checklist: AI in HR in 30 Days

Total: 12 hours over 30 days to transform HR from operational to strategic.

What to Test First

The paradox of AI in HR: the more you automate the operational tasks, the more time you have for what truly matters — understanding people.

The recruiter who spent 4 hours reading resumes now spends 1 hour having deeper conversations with candidates. The manager who used to improvise feedback now has data to give constructive input. The HR team that was always putting out fires now prevents problems.

AI doesn't replace the empathy of HR professionals. It creates space for empathy to happen.

The companies that will retain the best talent in 2026 aren't the ones paying the most. They're the ones treating people better. And treating people better requires time, attention, and data. AI delivers all three.

Want to apply AI in HR with practical lessons and ready-to-use templates?

• Explore AI courses for people management
• Access ready-made prompts for HR and recruiting
• See the complete guide to AI for HR professionals
• Create your free account and start today

The best HR department isn't the one that hires fastest. It's the one that hires better, develops people, and creates a place where talent wants to stay.