AI for Recruiters Playbook

How to Use This Template

Fill [[Tokens]] with your real details. Treat the prompt library and checklists as starting templates. Customize heavily for your roles, locations, and ATS before using with candidates. Always keep a human in the loop on final decisions.

What to Provide

Section 1: Use Cases by Hiring Stage

Sourcing and attraction

• Generate job description variants from a core spec for different channels.

• Draft outreach messages for LinkedIn, email, and communities.

• Identify likely skill synonyms and alternate titles for boolean searches.

Screening and evaluation

• Summarize resumes against must-have and nice-to-have criteria.

• Generate initial screening questions tied to the JD.

• Score structured responses using a consistent rubric.

Interview and selection

• Produce role-specific interview question banks with scoring anchors.

• Create scorecards and debrief templates.

• Summarize interview notes into strengths, risks, and recommendation options.

Offer and onboarding

• Draft personalized offer letters and onboarding plans (with legal review).

• Generate 30-60-90 day plans tailored to the role.

Your top priority stages to automate first:

1. [[Stage with biggest time sink]]

1. [[Stage with repeatable output]]

1. [[Stage where consistency matters most]]

Section 2: Prompt Library by Stage

Job description from notes:

"You are an experienced technical recruiter. Turn these notes into a clear, inclusive job description for [[Role]] at [[Company Stage/Size]]. Include: 3-4 impact areas, must-have vs nice-to-have in bullets, what success looks like in 6 months, and a short 'about us' paragraph. Use [[inclusive language rules]]. Output in clean markdown."

LinkedIn outreach (personalized):

"Write a concise, friendly LinkedIn message to [[Candidate Name]] who works at [[Current Company]] as [[Title]]. We are hiring a [[Role]] and their background in [[relevant thing from profile]] looks relevant. Mention one specific reason we think they would be a fit. Keep under 80 words. No hard sell."

Resume screen summary:

"Review this resume against the JD. Output: 1) Match score 1-5 on must-haves with one-sentence evidence. 2) Top 2 strengths for this role. 3) Top 2 risks or gaps. 4) Recommended next step (screen / reject / fast track). JD must-haves: [[paste list]] Resume: [[paste]]"

Interview question generator:

"Create 6 behavioral + 4 technical or work-sample questions for a [[Role]] interview. For each behavioral question add a 'what good looks like' 3-bullet scoring guide and a 'red flag' example. Base questions on these competencies: [[list 4-5]]."

Scorecard from notes:

"Turn these interview notes into a structured scorecard. Rate each competency 1-5 with one supporting quote from the conversation. End with overall recommendation and key open questions for the debrief."

Section 3: Tool Picks and Workflow Integration

Common stack layers:

• ATS core: Greenhouse, Lever, Ashby, Workable: use for structured data and workflows.

• AI layer: Many ATS now have native AI; otherwise use Claude or GPT with copy/paste or light automation.

• Sourcing: LinkedIn Recruiter + AI for message variants; SeekOut or similar for enrichment.

• Scheduling: Calendly or native with smart availability.

• Knowledge: Notion or Google Drive for JDs, rubrics, and past winning messages.

Integration pattern (low code first):

1. New application in ATS triggers a summary prompt via automation.

1. Output saved as a note or custom field on the candidate record.

1. Recruiter reviews summary + original before moving stage.

1. Approved questions or messages are generated on demand from templates.

When to go deeper: Once volume justifies, connect via Make/Zapier or native integrations to reduce copy/paste.

Section 4: Bias, Compliance, and Adverse Impact

Mandatory rules (do not skip):

• AI may assist with drafting and summarizing but final screening, interview, and selection decisions must involve at least one human who reviews the original materials.

• Never use AI to automatically reject candidates based on protected characteristics or proxies.

• Document the criteria you are using. If the AI is scoring, the rubric must be written and reviewed for disparate impact.

• For roles subject to EEOC or similar rules, keep records of how AI outputs were used and any overrides.

Bias check questions before using a new prompt or tool:

1. Could any language in the prompt systematically favor or disfavor a group?

1. Are we scoring on criteria that are actually job-related and consistently applied?

1. Have we tested the prompt on a diverse set of synthetic or historical profiles?

1. Is there a clear human override path and audit log?

Adverse impact monitoring: Track pass rates by stage for different demographic groups (where you have the data) at least quarterly. Investigate any large unexplained differences.

Section 5: Candidate Data Privacy

What you can safely send to consumer AI tools:

• Publicly posted profile information the candidate chose to share.

• Anonymized or redacted summaries when possible.

What you must not send (or must use enterprise zero-data-retention versions for):

• Full resumes with contact info, addresses, or IDs unless the tool contract allows it and you have consent language.

• Interview notes containing sensitive personal details.

• Any data from candidates in jurisdictions with strict rules (GDPR, etc.) without proper legal basis and processor agreement.

Simple habit: Before pasting a full resume, remove name + contact + location if not required for the task, or use an enterprise account with data protection terms.

Section 6: Structured Interview and Rubric Outputs

Always use structured interviews for roles with multiple candidates.

Required elements in every scorecard:

• Competencies tied directly to the job description.

• 1-5 or 1-4 scale with behavioral anchors.

• Space for evidence quote or example from the conversation.

• Overall recommendation with required calibration.

Example competency with anchors (adapt per role):

Collaboration (1-5)

5: Consistently elevates the team; gives specific examples of driving outcomes through others and handling conflict constructively.

3: Solid contributor in teams; shares credit and communicates clearly.

1: Works in isolation or creates friction; no clear examples of successful collaboration.

Section 7: ATS Workflow Tips

Recommended fields to add if missing:

• AI summary (text)

• AI match score (number 1-5)

• Human override reason (text, required when changing AI-assisted stage move)

• Source of AI assistance (which prompt / tool version)

Stage automation examples:

• Application received → AI summary generated and attached (human reviews before screen).

• Screen scheduled → Generate tailored screening questions and save to candidate.

• After interview feedback submitted → AI creates draft scorecard from notes for human editing.

Pro Tips

• Save every prompt that produces consistently good output in a shared library with version and role tags.

• Review AI summaries against the actual resume or conversation for the first 50 uses of any new prompt.

• Use AI to increase the number of strong candidates you can consider, not to shrink the funnel faster.

• Hiring managers love good question banks and scorecards: deliver those as your first AI win.

Common Pitfalls

• Letting AI write the entire JD without a human who owns the actual requirements.

• Using generic prompts that ignore role-specific competencies.

• Skipping bias testing because "the AI is neutral."

• Sending full personal data to consumer tools without review of the terms.

• Treating AI scores as the decision instead of one input to a human process.

AI can dramatically increase the quality and speed of your sourcing and preparation work. It cannot and should not replace human judgment on who gets hired. Keep the human accountable, keep the criteria explicit, and keep improving the inputs you give the AI. The best recruiting teams use AI to do more thoughtful hiring, not just faster hiring.