When somebody applies for a job on FindMeAJob, the employer does not get a bare CV in their inbox. They get a one page AI hiring brief alongside it. This is how it works, and what we do to keep bias out of it.
What the brief actually is
The brief is generated the moment an application comes in, from two inputs only: the candidate's CV and the employer's job description. It organises what is already there into a fixed structure. A recommendation, which is Interview, Maybe, Skip, or Insufficient data. A two sentence read on the candidate against the role. Three reasons they are worth talking to. One or two things worth probing when you meet them. A skill match showing which requirements from the ad are evidenced on the CV and which are not. And two or three questions written for that specific candidate, so a busy owner can run a useful 15 minute phone call without preparing one.
It is included on every plan, including Free. Employers pay for volume and reach, never for the brief itself.
One thing the brief is not: a decision. Our terms are blunt about this. The brief is a shortlisting aid, employers agree not to use it as the sole basis for any hire or no hire decision, and they agree to meet the candidate before any offer. AI organises. People decide.
Why we take bias seriously
Because the record of AI in hiring is genuinely bad when nobody is watching it. Amazon built a CV screening engine and scrapped it in 2018 after finding it had taught itself to penalise the word women's, because it had learned from a decade of mostly male tech CVs. In 2024, University of Washington researchers ran more than three million comparisons through AI resume screeners and found they favoured white associated names 85% of the time and female associated names 11% of the time. And this is not new to AI. A landmark 2004 study found that identical CVs got 50% more callbacks when the name at the top sounded white.
The pattern in all of it is the same: the damage comes from what a screener infers from things that have nothing to do with the job. A name. An address. A gap.
What we do about it
The AI is told what it is not allowed to use. Every single brief request carries mandatory fairness rules. The AI must not infer or use ethnicity, country of origin, age, gender, marital or family status, religion, disability, or sexual orientation. The candidate's name is passed for reference only, with an explicit instruction not to infer anything from it. There is no photo. Address based judgements are off the table.
Career gaps are never penalised. Parental leave, caregiving, study breaks, redundancy, sabbaticals, migration gaps and non linear paths are common in real NZ working lives, and the rules say plainly that none of them count against a candidate. Overseas experience is treated as experience. And the AI never rates anyone's English: typos or awkward phrasing on a CV are a coachable formatting note, never a proxy for where somebody comes from.
Every applicant gets the same treatment. Same structure, same job ad, same rules, for every candidate on a role. Research on hiring consistently shows that structured, criteria based evaluation reduces bias compared with gut feel, and the brief is structured evaluation by design. We also run the AI at a deliberately low creativity setting, so the same CV against the same ad gets a consistent read rather than a different verdict depending on the day.
Entry level is judged as entry level. This board exists for people with little or no paid history, so the AI is instructed to mine CVs for transferable evidence, such as study, sport, volunteering, caregiving and community roles, and to treat a lack of paid experience as never, on its own, a reason to skip someone. When in doubt, it is told to lean towards Interview.
The honest part
Large language models are probabilistic, so we do not claim these rules hold perfectly in every single case, and our terms say so in writing. What we do instead is keep humans in the loop at both ends. Employers must meet the candidate before any offer. And any candidate who believes a brief misrepresents them can dispute it by emailing hello@findmeajob.co.nz. We review the brief and its inputs, respond within 20 working days, and can ask the employer to pause the application while we look. We keep an audit log of sensitive operations for 12 months, and all of it sits under the NZ Human Rights Act 1993 and the Privacy Act 2020, in line with the Privacy Commissioner's AI guidance.
The full detail lives on our Trust and AI Fairness page, and the binding versions are in our terms and privacy policy. If you read one line of all this, make it this one: the AI reads, the human hires.
Frequently asked questions
Is the AI brief a hiring decision?
No. It is a shortlisting aid. Employers on FindMeAJob agree not to use it as the sole basis for any hire or no hire decision, and to meet the candidate before any offer. The AI organises the application; the employer decides.
What is the AI told to ignore?
Ethnicity, country of origin, age, gender, marital or family status, religion, disability and sexual orientation. It is also told never to penalise career gaps and never to rate English language ability. The candidate's name is passed for reference only, with an instruction not to infer anything from it.
What does the brief cost an employer?
Nothing. The AI brief comes with every applicant on every plan, including Free. Paid plans add listing volume and reach, not the brief.
What if a brief gets me wrong?
Email hello@findmeajob.co.nz with your application reference. We review the brief and the inputs it was generated from and respond within 20 working days, and we can ask the employer to pause the application while we look.