JD Creator
Job descriptions drafted from scratch
Give it the role details. Get a clean, copy-paste-ready posting that sounds like your company actually wrote it, not a generic template.
The 5-phase process
1
Gathers context
Role, company, responsibilities, must-haves vs. nice-to-haves, logistics, compensation, and what makes the role distinctive. Never invents data: if something's missing, it asks.
2
Confirms structural fields
Title, seniority level, contract type, experience, location, work model, languages, salary band.
3
Sets the tone
Chooses the voice that fits the company: corporate, founder voice, or confidential.
4
Drafts in 4 blocks
Clean, structured output ready to paste anywhere.
5
Iterates
Refines until you're happy, then delivers the final text.
The 4 blocks of the posting
A
About the company / role vision
2-3 paragraphs, strongest points first.
B
What you'll do
Responsibilities in action-verb bullets.
C
What we're looking for
Requirements with must-haves and nice-to-haves clearly separated.
D
What we offer
Real compensation and benefits.
Three tones available
Corporate / professional
Founder voice
Confidential search
Comes with 2 built-in style examples. Paste up to 3 of your own previous job postings and they take priority, so the output sounds like your company, not a template.
Built-in guardrails
📊
Leads with concrete data
Numbers and real facts over buzzwords.
🎯
Keeps requirements tight
Pushes back if you inflate must-haves, for a bigger and fairer candidate pool.
🌍
Inclusive language
No native-speaker bias, age/gender markers, or self-excluding requirements.
✅
Confirms before inventing
Never fabricates company figures, salary, or requirements.
💰
Salary transparency
Recommends publishing the band and reminds you to check local requirements.
Killer Questions
Knockout questions for screening calls
Prepares the knockout questions for a pre-screening call: fast filters against the role's non-negotiables, not an in-depth interview guide.
- Confirms the role's non-negotiables with you before generating a single question
- Each question is open-ended with a 3-way route: Dig in / Pass / Drop
- Weighted as [KNOCKOUT] or [STRONG SIGNAL], never a blind auto-reject
- Built-in bias scan reframes any protected-characteristic criteria before they become a question
- Follows a natural call arc, and can wrap into a 1-page cheat sheet for the call
How it builds the list
1
Gathers the material
Reads the JD and briefing from a connected ATS or from what you paste, or asks a few quick calibration questions if you have neither.
2
Confirms the non-negotiables
Shows the knockouts and must-haves it detected and waits for your confirmation: a mandatory checkpoint before it generates anything.
3
Delivers the call-ready list
Each question with its weight and its 3-way route, ordered as a natural call arc, plus an optional 1-page sheet to mark pass/fail live.
Built-in guardrails
- Every criterion must be job-relevant, protected characteristics get reframed, never turned into a question
- Same bar for every candidate in the process, no exceptions
- A knockout is a signal to explore, not an automatic rejection
Candidate Messages
Outreach and rejection messages
Drafts the two candidate messages recruiters get wrong most: attracting a profile to a role, and saying no gracefully, with optional constructive feedback.
- Detects whether you need outreach or a rejection, and asks only what's specific to that message
- Outreach: role-tailored hook, a single clear CTA (book / reply / phone), plus an optional follow-up for non-responders
- Rejection: calibrated by process stage (applied, interviewed, or final) as neutral, reason-based, or with real constructive feedback
- Built-in bias scan on both flows: a protected-characteristic reason never becomes a written rejection
- Your profile (name, company, in-house vs. agency) is saved once and reused for every message
Two flows, one skill
1
Save your profile once
Name, role, company, and whether you recruit in-house or for an agency client, never asked again.
2
Tell it outreach or rejection
Outreach asks for the JD and the goal of the message; a rejection asks for the process stage and whether you want it neutral, reason-based, or with feedback.
3
Delivers it ready to copy
Outreach comes with an optional follow-up for non-responders; a rejection is calibrated in tone and warmth to the stage the candidate reached.
Built-in guardrails
- Reason and feedback are always about the role, never personal characteristics
- If the real rejection reason is a protected characteristic, only the neutral no-reason version is offered
- Drafts only: you review and send, nothing is ever sent automatically
CV Screening
Traffic-light scoring for every CV
Screen a batch of CVs against a role's real criteria and get a decision table: traffic light, fit %, and evidence for every candidate.
- Sorts criteria into knockouts, must-haves, and nice-to-haves, confirmed with you before screening starts
- Fixed 80/20 fit-score formula: must-haves carry the weight, nice-to-haves only break ties
- 🟢 Advance / 🟡 Maybe / 🔴 Reject for every CV, backed by evidence cited from the resume
- No evidence is flagged "to confirm in interview," never scored as a silent reject
- Two-layer output: a compact summary table always, detailed cards only for who you choose
How it screens
1
Gathers CVs and criteria
Reads CVs from a connected ATS, a folder, or pasted directly, then extracts and confirms the knockouts, must-haves and nice-to-haves with you.
2
Scores every CV
Knockouts first, then each must-have and nice-to-have. Every score is backed by evidence cited straight from the resume, never a guess.
3
Delivers the shortlist
A sortable summary table for everyone, ranked by fit %, with optional detailed cards for the candidates you pick and a one-click CSV export.
Built-in guardrails
- Scores only what's job-relevant: age, photo, origin, and similar never enter the evaluation
- Company or university pedigree can add as a nice-to-have, but never acts as a knockout
- The skill recommends; the final decision is always the recruiter's
Recruiter Assistant
Market intel to interview debrief
One skill, three moments of the hiring process: briefs the role, builds the interview guide, and scores the interview afterward.
- 7-block market briefing: day-to-day, education, up to 10 target companies, LinkedIn boolean, salary table, similar titles, and a quick interview script
- In-depth interview guide, with a desirable answer and red flags for every technical and behavioral (STAR) question
- Post-interview debrief: scores the candidate against the JD with a qualitative rating and an advance/reject call
- Company & salary mapping is built in, no separate skill needed
Three capabilities in one skill
1
TA Expert: market briefing
Give it a job title (+ country). Get day-to-day tasks, typical education, up to 10 target companies, a LinkedIn boolean, a salary table by experience band with sources, 5 similar titles, and a quick interview script to kick off with.
2
Interview Guide
Builds the full guide for that role: technical questions with a real scenario each, and STAR behavioral questions, every one paired with a desirable answer and red flags to watch for.
3
Debrief
Paste the transcript or notes afterward. It scores each competency against the JD/checklist (🟢🟡🟠🔴⚪) and gives an Advance / Advance with reservations / Reject call.
Built-in guardrails
- Judges only what's relevant to the role, never accent, age, gender, origin or appearance
- Labels anything unverified as "Estimate (unverified)" instead of presenting it as fact
- Never persists candidate names, notes or transcripts, only the role's process record
Boolean Search Builder
Ready-to-paste sourcing strings
Picks the right output for your tool: Boolean strings for LinkedIn Recruiter, or a full search JSON for Nova Recruiter.
- Asks your sourcing platform first: LinkedIn/Sales Navigator gets Boolean, Nova Recruiter gets ready-to-paste JSON
- 2-3 calibrated variants: broad, balanced, and precise, each with a note on what it catches and misses
- Synonyms and title variants in the local market language + English: the biggest coverage multiplier
- Confirms role, location, experience, industries, skills and exclusions before generating anything
Two branches, same calibration step
1
Asks your platform first
LinkedIn Recruiter / Sales Navigator, or Nova Recruiter. The output format depends on it, so it's the first question, always.
2
Confirms the criteria
Role and titles, location and radius, experience range, industries, must-have skills, languages and exclusions, confirmed before generating anything.
3
Generates the search
LinkedIn: 🔵 Broad, 🟢 Balanced, 🔴 Precise Boolean strings, each with a note on what it catches and misses.
Nova: the full search JSON (keywords, location, experience, companies, languages, education) ready to paste into the agent.
Built-in guardrails
- Synonyms always in the local market language + English: the biggest coverage multiplier
- Never invents filters or exclusions you didn't confirm
- Nova's exact industry/subindustry catalog is used so filters don't silently get dropped
Outreach Messages
Multi-channel cold outreach
One brief, three ready-to-send formats: LinkedIn connection note, LinkedIn message/InMail, and email, each with a follow-up for non-responders.
- Each channel in its own format: ultra-short connection note, a real pitch for LinkedIn/InMail, subject + body for email
- Built-in bias scan on the JD and the hook before writing a single word
- Follow-up sequence with a new angle plus a graceful "not for you? refer someone" exit
- Reusable
How a role gets messaged
1
Your profile, once
Name, title, company, and whether you recruit in-house or for an agency client, saved once and reused for every role.
2
The role brief
Paste the JD (or the 2-3 things that make it attractive), the goal of the message (book, reply, or phone number), and which channels you want.
3
Three channel-native drafts
An ultra-short connection note, a LinkedIn message/InMail with subject, and a full email, each with one follow-up carrying a new angle and a graceful referral exit.
Built-in guardrails
- Scans the JD and hook for biased language before writing a single word
- Never promises salary or conditions that weren't confirmed
- Drafts only: you review and send, nothing is ever sent automatically
Scoring Model
Weighted competency rubric for interviews
Builds the evaluation framework for a role's interviews: 4-6 weighted competencies with behavioral anchors per level, plus a scorecard ready for the debrief, so every interviewer scores by the same yardstick.
- Confirms role, seniority and must-haves with you before proposing a single competency
- Max 6 competencies, technical + behavioral blocks, weights in % that must sum to 100
- 4 behavioral anchors per competency (1-4), calibrated to seniority, describes what "meets" looks like, not adjectives
- Marks [CRITICAL] competencies with a minimum acceptable level: falling below flags for review, never auto-rejects
- Can also fill in a specific candidate's scorecard if you provide their notes or transcript (2-in-1 mode)
How it builds the rubric
1
Gathers the role
Reads the JD/briefing (or a connected ATS) and confirms its understanding of the role before proposing anything.
2
Proposes the skeleton
Competencies, weights and criticals in a table, and stops for your confirmation before writing a single anchor.
3
Writes the full rubric
4 anchors per competency plus a blank scorecard ready for the debrief, or filled in if you're evaluating someone.
Built-in guardrails
- Only competencies relevant to the role, never age, origin, accent, appearance or personal life
- "Culture fit" only enters operationalized as observable behavior, never raw affinity
- The skill never rejects on its own: a critical flag forces human review, doesn't decide
Interview Evaluation
Scorecards and recommendations from evidence
Scores one or several candidates against an evaluation rubric, straight from their interview notes or transcript. Returns a scorecard per candidate, weighted score, cited evidence, flags and a recommendation, plus a comparison table if there are several.
- Scores only what the notes evidence; no signal is marked N/A, never a 1
- Same rubric and thresholds for every candidate, no softening or hardening the bar
- Flags any critical competency below the minimum, but never rejects on its own
- Surfaces an "extra potential" axis for standout strengths outside the rubric, without mixing it into the score
- Builds you a rubric first if you don't already have one
How it evaluates
1
Checks for a rubric
Uses yours if you have one, or builds one first (chains into Scoring Model if it's installed).
2
Scores each candidate
Competency by competency, with cited evidence, weighted into a score out of the rubric's scale.
3
Delivers the scorecard(s)
One per candidate, plus a ranked comparison table for multi-candidate rounds.
Built-in guardrails
- Scans any rubric you bring for bias proxies ("culture fit", "digital native"...) before applying it
- Never scores by protected characteristics; evidence over interviewer opinion, always
- Recommendation, never a verdict: "Advance / with reservations / Don't advance", the decision stays yours
Interview Summary
Structured, honest interview write-ups
Turns your interview notes or transcript into a clean, structured summary ready to share with the hiring team or store in the ATS, organized, evidence-based, and without scoring or a hire/no-hire verdict.
- Organizes notes into Background, Strengths, Development areas, Motivation and Key quotes
- Flags gaps in coverage before generating, so you can add anything from memory first
- Never infers a weakness from silence: what wasn't probed goes to "To confirm", not "Development areas"
- Unpacks unsupported labels like "culture fit" or "not a fit" instead of copying them as fact
- Preserves contradictions in the notes rather than resolving them in favor of one side
How it builds the summary
1
Gathers the material
Notes, transcript and JD if you have them, plus any non-negotiable you want it watching for.
2
Confirms and flags gaps
Reflects back what it understood and tells you what has no signal before generating.
3
Delivers the write-up
A fixed, readable structure, ready to paste into an email, doc or ATS.
Built-in guardrails
- Only what's in the notes; a section with no signal is marked "not covered", never invented
- Personal or protected-characteristic disclosures never carry into quotes or the summary
- It's a record, not a verdict: no scoring or hire/reject recommendation
Interviewer Feedback
Coaching on interview technique
Gives constructive feedback on HOW an interview was conducted, your own or a teammate's, from the transcript. Coaching by dimension, always anchored to a verbatim quote, with rewritten questions and 2-3 quick wins.
- Every claim carries a transcript quote; without one, the claim is dropped
- Judges a moment, not a pattern: never "you tend to" or "you're too..."
- Prioritizes the 2-3 dimensions with the most to gain, not a forced 7-point review
- Bias only flagged with a quote behind it; may say "no risk observed" and mean it
- Refuses to evaluate the candidate, that's a different skill, and it says so
How it coaches
1
Confirms who and what
Whose interview it is, and who's speaking in the transcript, before analyzing anything.
2
Analyzes by dimension
Question quality, probing, coverage, signal extracted, gaps and bias, only with cited evidence.
3
Delivers the coaching
What worked, what to improve, and the rewritten version of the question, plus quick wins.
Built-in guardrails
- Coaching, not a performance record, meant to be shared with whoever interviewed, not filed against them
- Personal style (pace, accent, humor, silences) is never a flaw unless it costs signal, and that's cited
- Focus stays on the interviewer's conduct, never a verdict on the candidate
Nova Recruiter
