How to start, what to type, and how to get value fast—without risking candidate data or introducing bias.

1) Start Here: The 3-Step Method (Hide → Ask → Check)

Recruitment data is sensitive (GDPR/privacy is critical). Treat AI like a coordinator: you own the relationship; it handles the paperwork.

1. Hide candidate info (Anonymise)
2. Ask for a specific outcome
3. Check for bias and accuracy
   • Did it invent qualifications?
   • Is the language inclusive?
   • Replace names, emails, phone numbers, and companies with tokens:
     • Jane Doe, Google, 07700900 → [CANDIDATE], [CURRENT_EMPLOYER], [PHONE]

One rule: No CVs uploaded directly. No names. No salary slips.

2) What to Type: Prompt Anatomy (copy this pattern)

Role/Task + Context + Constraints + Format

You are a Technical Recruiter.

Task: Draft a cold outreach message to a passive candidate based on the redacted profile summary below.

Context: We are hiring a Senior Java Developer. The candidate has 5 years of experience in Fintech but hasn't updated their profile in 6 months.

Constraints: Under 100 words, friendly but professional tone, mention our "remote-first" policy. No generic fluff.

Format: Subject line options + Body text.

3) First 10 Prompts to Try (beginner-safe)

1. Boolean String Builder

"Write a Boolean search string for LinkedIn Recruiter to find [JOB TITLE] in [LOCATION]. Must have [SKILL 1] and [SKILL 2]. Exclude [COMPETITOR NAMES]."

2. Job Description (JD) Refinement

"Rewrite this list of requirements to be more inclusive and appealing to diverse candidates. Remove gender-coded language and focus on outcomes rather than years of experience.

Draft: [PASTE BULLETS]"

3. Interview Question Generator

"Based on the redacted job description below, generate 5 behavioral interview questions that test for [SPECIFIC SOFT SKILL, e.g., adaptability].

JD: [REDACTED TEXT]"

4. Candidate Summary (for Hiring Managers)

"Summarise these redacted interview notes into a 'Candidate Snapshot' for the hiring manager. Format: Key Strengths, Potential Gaps, and Recommended Level.

Notes: [REDACTED NOTES]"

5. Cold Outreach (Hyper-Personalized)

"Draft a 3-sentence hook for a LinkedIn message connecting [CANDIDATE'S INTEREST IN AI] to our company's new project on [PROJECT TOPIC].

Tone: Casual professional."

6. Rejection Email (Empathetic)

"Write a rejection email for a candidate who reached the final stage. Acknowledge their strong [SKILL] but explain we went with someone with more [OTHER SKILL] experience. Keep it warm and encourage future applications."

7. "Explain Like I'm 5" (Tech Screening)

"I am a non-technical recruiter. Explain the difference between 'Java' and 'JavaScript' in simple terms so I can screen candidates accurately. Give me 3 'red flag' answers to watch out for."

8. Offer Letter Helper

"Draft the 'Benefits' section of an offer letter using these raw notes. Make it sound exciting, not legalistic.

Notes: [25 days holiday, private health, gym budget...]"

9. Competitor Research

"List the top 5 typical pain points for employees working in [SPECIFIC INDUSTRY/ROLE]. I want to address these in our value proposition."

10. Bias Check (Linting)

"Scan this feedback draft for unconscious bias or subjective language (e.g., 'culture fit', 'articulate'). Suggest more objective alternatives based on evidence.

Text: [DRAFT FEEDBACK]"

4) Live Workflow (Sourcing & Screening)

Sourcing:

• Use Prompt #1 to build search strings.
• Paste redacted LinkedIn summaries to generate personalized hooks (Prompt #5).

During Screening:

• Type notes with tokens ([CANDIDATE], [ROLE]).
• Mid-call helper: "What is a polite way to ask about notice periods and salary expectations without sounding aggressive?"

Post-Interview:

• Run Prompt #4 to clean up your notes before sending them to the Hiring Manager. Check against your actual memory of the call.

5) Before/After (see the difference)

❌ Unsafe

"Write a summary for Sarah Jones, born 12/05/1990, lives at 42 Wallaby Way. She worked at CompetitorX."

(Risk: PII leakage, data retention issues)

✅ Safe (after Hide → Ask → Check)

"Write a summary for a candidate with 5 years of experience in high-volume retail.

Key achievement: managed a £2M budget.

Gap: lacks strategic planning experience.

Format: Bullet points for the hiring manager."

6) Common Mistakes (and quick fixes)

• Pasting CVs → Never do this. Copy sections of text and remove headers/footers first.
• "Is this candidate good?" → AI cannot judge character. Ask it to compare skills against a list, not to make a judgment call.
• Leaving in biases → If you feed it biased notes, it gives biased summaries. Always use the "Bias Check" prompt.
• Generic Outreach → If you don't give specific context, your emails will sound like spam.

7) What AI Is Great At vs. Not Great At

Great at:

• Parsing messy notes into structured summaries.
• Writing Boolean strings.
• Checking JDs for exclusive language.
• Drafting repetitive emails (rejections, scheduling).

Not great at:

• Assessing "Culture Fit."
• Verifying truth (it can't do background checks).
• Understanding nuance in gaps on a CV.

8) Mini-Checklist (stick near your screen)

• [ ] I removed Name/Phone/Email/Address.
• [ ] I removed current employer names if confidential.
• [ ] The prompt focuses on skills and evidence, not "vibes."
• [ ] I verified the Boolean string works before sending it.
• [ ] I checked the final text for American vs. British English.

9) Tools & Setup

• Microsoft Copilot (Enterprise) or ChatGPT Enterprise (ensure 'Training' is off).
• ATS Integration: If your ATS has AI built-in, use that for data within the system, but use external AI only for drafting text.

10) 15-Minute Team Starter

1. 5 min: Show this guide. Pick a real (old) Job Description.
2. 7 min: Team uses Prompt #2 (JD Refinement) and Prompt #1 (Boolean) to improve the search for that role.
3. 3 min: Share the best boolean string or JD rewrite. Save it to the team "Prompt Library."