Screening support
Structured assessment of every application against the documented criteria for the role, with the evidence shown per criterion - so a recruiter reviews reasoning rather than skimming CVs.
Industries
Recruitment AI that will not fail a tribunal
Screening is the highest-volume task in recruitment and the single most legally exposed use of AI in any organisation. It can be done well, but only with bias testing, a human decision and a record of how each candidate was assessed.
The pressure right now
What we hear from recruitment & hr
- Application volumes per role that no team can genuinely read
- Time-to-hire losing you candidates to faster competitors
- Screening consistency varying by recruiter, mood and hour of the day
- Candidate experience suffering from silence and generic rejections
- HR case administration and employee queries absorbing the team's week
Where it pays off
Six places AI earns its keep here
Not everything on this list will apply to you. Most organisations start with one and extend once it has been measured.
Scheduling and coordination
The interview scheduling, rescheduling and reminder loop, which consumes an astonishing share of recruiter time and requires no judgement at all.
Job content and adverts
Role descriptions and adverts drafted from your framework, checked for exclusionary or gendered language before they are published.
Candidate communication
Timely, specific updates and rejections at volume - the change candidates notice most, and the cheapest reputation improvement available.
Employee query handling
Grounded answers from your own policies and handbook for the routine questions, escalating anything about a person's circumstances straight to HR.
Case administration
Assembling the chronology and documentation for grievance, absence and performance cases, with the judgement and the decision staying with HR.
Where we would start
The first three moves
1
Start with scheduling, not screening
It is pure time recovery with no legal exposure, and it funds the more careful work that follows.
2
Document the criteria before automating them
Most screening bias comes from criteria that were never written down. Writing them is half the fix.
3
Test for adverse impact
Before any screening support goes live, and on a schedule afterwards, with the results recorded.
Risk and regulation
The part most suppliers skip
Where the risk sits
- Equality Act 2010: indirect discrimination through a proxy the model learned
- UK GDPR rights around solely automated decisions with legal or similarly significant effects
- Employment-related AI is classified as high risk under the EU AI Act where it applies
- Candidate personal data retention and the right to an explanation
- Historic hiring data encoding historic bias into a screening model
How we handle it
No candidate is rejected by a machine. The system assesses and evidences against documented criteria; a person makes and records the decision, which also keeps you outside the automated-decision restrictions.
Adverse impact testing across protected characteristics happens before launch and on a schedule afterwards, with the results written down and reviewable.
We avoid training on your historic hiring outcomes, because that is precisely how a model learns yesterday's bias and launders it as objectivity.
Where this fits
Where to go next
Questions
Questions from recruitment & hr
We will not build that. Beyond the discrimination exposure, a solely automated rejection with significant effect engages specific data protection rights. Keeping a human decision is both safer and, at realistic volumes, barely slower.
By testing outcome rates across protected characteristics on your own data before launch and periodically after, and by keeping the assessment criterion-based and explainable rather than a similarity score against past hires.
Yes - we recommend saying so plainly in your privacy notice and application process. Transparency is a legal expectation and, in practice, candidates mind far less than teams fear when a human decision is guaranteed.
Start with an audit of what you already run
Two to four weeks to an evidenced picture of your AI use, spend and risk - and a ranked list of what to do first.