AI resume screening that assists recruiters, and never rejects anyone
The first pass through a stack of CVs is the least glamorous part of hiring and the biggest time sink. A recruiter reads the same fields off every application, tries to hold a consistent bar across a hundred documents, and loses hours to formatting instead of judgement. AI is genuinely good at that reading and organising. What it must never do is decide who is out. Roiwerk builds screening automation that structures and ranks applications against your actual criteria, explains its reasoning, and hands a clean shortlist to a human who makes every call.
What the automation actually does
When an application arrives, the workflow pulls the CV and any attachments, extracts the details that matter for the role, and normalises them into a consistent structure regardless of how the candidate formatted their document. An LLM reads the free text the way a recruiter would, mapping experience, skills, and history against the requirements you defined for that specific job. The output is a tidy summary per candidate: what the role asked for, what the application shows, and where the gaps are.
Crucially, the ranking is a sorting aid, not a verdict. Candidates are grouped and ordered so the recruiter starts with the strongest matches, but every application stays visible and reviewable, and the reasoning behind each placement is written out in plain language. Nobody is filtered into a bin the recruiter never opens. The point is to change where the recruiter spends their attention, not to make the decision for them.
- Parse CVs and attachments into a consistent structure, whatever the format
- Map experience and skills against the requirements you set for the role
- Produce a plain-language summary of fit and gaps per candidate
- Rank as a sorting aid, with every application still visible and reviewable
- Write the reasoning behind each placement so a recruiter can check it
Why we will not build an auto-rejecter
The obvious next step, letting the model reject the bottom of the pile automatically, is the one we refuse. A screening model trained or prompted on past hiring can absorb and amplify bias, disadvantaging candidates for reasons that have nothing to do with the job. Auto-rejection also strips away the human review that catches the model's mistakes: the strong candidate with an unusual career path, the person whose CV undersells them, the false negative no one ever sees.
There is a legal dimension too. Under the GDPR and related rules, decisions with a significant effect on a person that are made solely by automated means are tightly restricted, and hiring is squarely in that territory. So we design for the opposite: the AI proposes an ordering and an explanation, a person reviews it, and the human decision is the one that counts. That is not a compliance afterthought; it is how the system earns trust.
- No candidate is ever rejected by the model; humans decide who advances
- Human review catches false negatives a scorer would silently drop
- Criteria are explicit and documented, not hidden inside a black box
- Designed to respect GDPR limits on solely-automated hiring decisions
Keeping it fair and auditable
Screening against the right things is how you keep it fair. We work with you to turn a role into explicit, job-related criteria before anything runs, so the AI is measuring what the job needs rather than pattern-matching on prestige or proxies. Where it helps, we steer the model away from fields that invite bias and toward evidence of the actual skills and experience the role requires.
Everything the automation does is logged and explainable. For any candidate, you can see what the model read, how it mapped to the criteria, and why it landed where it did. That audit trail protects candidates from arbitrary treatment and protects you if a decision is ever questioned. It also lets you tune the criteria over time as you learn what a strong hire for the role really looks like.
Built into your hiring stack
The screening layer connects to the tools you already use: your ATS or the inbox and forms applications arrive through, and your shortlist and notes flow back to wherever your recruiters work. Summaries land on the candidate record, so a recruiter opens one place and sees the structured read alongside the original CV. No new tool to learn, no data marooned in a system nobody checks.
As with everything we build, it runs in your accounts, the logic is documented, and you own it. If your requirements change or you want to adjust how strictly the model reads a criterion, your team can do it. We would rather leave you with a screening process you understand and control than a scoring engine you have to take on faith.
- →AI does the reading and organising; recruiters make every decision about who advances.
- →No auto-rejection, ever: ranking is a sorting aid with all applications visible and the reasoning shown.
- →Explicit, documented, job-related criteria and a full audit trail keep screening fair and defensible.
Does the AI reject candidates automatically?+
No. It reads, structures, and ranks applications against your criteria and explains its reasoning, but every application stays visible and a recruiter decides who advances. We do not build auto-rejection, both because it hides the model's mistakes and because solely-automated hiring decisions are legally restricted in Europe.
How do you stop the screening from being biased?+
We turn each role into explicit, job-related criteria before anything runs, steer the model toward evidence of required skills rather than proxies, and keep every ranking explainable and human-reviewed. The criteria are documented and the reasoning is logged, so you can audit any decision and tune the criteria as you learn.
Will it work with the way applications reach us today?+
Yes. Whether candidates come through an ATS, a careers form, or email with attachments, we connect to that source, structure what arrives, and write clean summaries back to the candidate record your recruiters already use. There is no new tool for your team to adopt.
Not sure which applies to you?
Book a free assessment and we'll map the highest-ROI automation opportunities for your business, honestly, including when it's not worth starting yet.
Book a free AI assessment