Is AI recruiting biased?
It can be, and the usual source is not the model but the data and the proxies. A system trained on who a company hired before will reproduce who that company hired before, and ranking on prestige signals like employer brand or school quietly encodes the same thing. What reduces it is ranking on evidence of work, keeping the scoring reproducible, and having a human make every decision that affects a person.
$100 of credits on us. No card required.
Our legal team will ask about this. What is the real answer?
'%20fill='%23FCFCFC'/%3e%3cellipse%20cx='304.949'%20cy='107.231'%20rx='43.7363'%20ry='86.7747'%20transform='rotate(-12.9533%20304.949%20107.231)'%20fill='%23FCFCFC'/%3e%3cellipse%20cx='207.296'%20cy='107.321'%20rx='33.7301'%20ry='67.712'%20transform='rotate(-12.2177%20207.296%20107.321)'%20fill='%23121212'/%3e%3cellipse%20cx='301.354'%20cy='103.848'%20rx='31.5494'%20ry='64.0532'%20transform='rotate(-12.2177%20301.354%20103.848)'%20fill='%23121212'/%3e%3c/svg%3e)
Honest version: any ranking system can encode bias. Mine is deterministic and shows its evidence, so you can inspect why someone ranked where they did.
And I never reject anyone. I rank and explain. Humans decide.
Humans make every decisionFour places bias actually enters
Training on past hires
If the target variable is who you hired before, the output is a mirror of your history. This is the most common and most damaging design choice.
Prestige proxies
Ranking on employer brand, school or title is ranking on access to opportunity. Judging shipped work is closer to judging the person.
Unstable scoring
If a score changes between runs, nobody can audit it. Reproducibility is a fairness property, not just an engineering one.
Automated rejection
The moment software rejects a person without a human reading the case, a small statistical error becomes a real harm at scale.
I score against the bar you describe, not against your hiring history, and every line of a score points at the artifact behind it.
The same brief produces the same reasoning and the same shortlist, so a decision can be reviewed months later.
I never reject, never send and never book anything without a person approving it. Sourcing on potential also tends to surface people the prestige filters would have skipped.
Recruiting judgement ships with me
You are not writing prompts to teach me what a strong candidate looks like. I already know what this job is, which is why the first shortlist is worth reading.
Search, screen, write, schedule and log are not five tools with five bills and five handoffs. They are one pass.
An independent score, not a claim I made up. Read how it was measured before you believe it.
Add me to a channel, describe the person you need, and the first shortlist arrives while a rollout plan would still be in review.
I do the volume. Who to meet, what to pay and who to hire stays with the humans whose names are on the offer.
The questions that usually come next
Does Laidback replace our ATS?
No. Your ATS stays the system of record. I work on top of it, read from it and write outcomes back so nobody does evening data entry.
Who sends the outreach?
Your team does. I draft messages from the evidence in each profile and a human approves and sends them from your own inbox.
How long until the first shortlist?
The same day. Connect your ATS and calendar, brief me in a channel, and the first evidenced shortlist comes back in that thread.
How this works with your ATS
See it on your own open role
One sentence in a channel and I will come back with an evidenced shortlist.