For hiring teams
AI hiring, without the AI slop
“AI hiring” usually means one of two things: software that reads résumés and rejects people, or software that writes recruiter spam at volume. We do neither. AI is useful here for matching and drafting; it is not useful for deciding who deserves a job.
How it works
AI ranks, humans decide
Models score how well a candidate fits a brief and surface people a keyword search would miss. Every shortlist that reaches you was reviewed by a person, and no candidate is ever rejected by a model.
Matching on stated preferences, not inferred ones
Candidates declare their salary floor, location, work mode and dealbreakers. Matching runs on what they told us, not on what a model guessed about them from a CV.
Drafting is assisted, sending is not automated
Outreach is drafted with AI and sent by a human against a hard cap of three companies per candidate per month. That cap is enforced in the database, not by policy.
Salary transparency is a data requirement
Because every role carries a real band, matching can rank on compensation fit instead of guessing. That is also what makes the aggregate salary data honest.
Questions
- Do you use AI to screen out candidates?
- No. Models are used to rank and surface candidates, never to reject them. Every rejection is a human decision, and a candidate who is not shortlisted is not silently filtered out of future searches.
- How is AI used in matching?
- It scores fit between a role brief and candidate profiles — skills, seniority, domain, stated preferences and compensation band — and surfaces strong matches that keyword search misses. A recruiter reviews the ranking before you see anything.
- Will candidates get AI-generated spam?
- No. Outreach may be drafted with AI, but it is reviewed and sent by a person, and every candidate is capped at three company approaches per month. The cap is enforced by a database trigger, so it cannot be bypassed by a campaign.
- Is my hiring data used to train models?
- Role briefs and candidate data are used to run your searches, not to train third-party foundation models.