AI CV screening, done defensibly, reads each application against the requirements a hiring manager wrote down and quotes where each one is or is not evidenced. It does not score, rank or reject anyone. A recruiter reads the evidence and decides.
Most AI resume screening tools produce a number per candidate and a cut-off. It is the easiest thing to build and the hardest thing to defend. A score compresses a dozen judgements into one figure nobody can unpack, it inherits whatever patterns were in past hiring, and a cut-off turns it into a decision about a person that no person made. Under the EU AI Act, AI that filters or evaluates applications is high-risk; under GDPR Article 22, decisions based solely on automated processing are restricted. A score with a threshold walks straight into both.
So this system answers a narrower question: for each requirement on the job specification, what in this application speaks to it? Five years of B2B sales: here is the line in the CV, and here are the dates. Fluent German: not mentioned. The recruiter sees every application, in an order they choose, with the evidence laid out. They decide, and the log records what they saw when they did.
Before go-live, we test the extraction on your own historical applications. The question is whether it finds evidence equally well for everyone: does it miss experience described in a non-native register, under-read career breaks, or treat foreign qualifications differently from local ones? We build matched test pairs (the same experience, written differently) and measure the gap. Where the data supports it, we look at results across the groups you lawfully monitor.
After go-live, the firm monitors outcomes: who progresses at each stage, by group, compared with who applied. That is the measure that matters, and it is the recruiter’s decisions being measured as much as the software’s. Candidates are told that AI assists the review, and how to ask for a person to look at their application again.
Where an employer offers that choice, it is usually a request for a person to review your application directly, and it is reasonable to use it if you think software may misread your experience. In the systems we build the question does not arise in the same way: no software rejects anyone, every application reaches a recruiter, and candidates can ask for a fresh human review.
Two ways, before and after launch. Before, we test the extraction with matched pairs of applications that describe the same experience differently, and measure whether evidence is found equally well. After, the employer monitors progression rates by group at each stage against the applicant pool. Bias can enter through the criteria, the software or the people, so all three are checked.
It is allowed, but AI used to analyse and filter applications or evaluate candidates is classed as high-risk. Deployers must assign competent human oversight, keep logs, use the system as instructed and inform candidates, and workers’ representatives where applicable. Whether your system is in scope and what follows is your legal team’s determination; we design so those duties can be met.
The extraction step reads the whole document, because it has to find the evidence, but the recruiter’s evidence view hides names, photos, dates of birth and addresses by default. That reduces the chance of a first impression being formed on irrelevant details, though it does not remove bias by itself, which is why outcomes are monitored.
What we built, what stage it is at, and a link to the live thing wherever there is one.
yourOS is an operating system with an AI at its heart. You talk to it like a person, out loud or typed. It makes your calls, plans your days and works on your projects with a crew of agents, on your own private server. Our own product, built from the first commit; the waitlist is open.
Website, backend and app assets for Sonora, a free AI sound-therapy app on iOS and Android published by Cognitive AI.
Brand, website, prototype, apps on five platforms, the AI backend and the algorithm-intelligence pipeline for Podswap, a social growth engine that runs without a human in the loop.
In closed beta: Roxy, a private AI assistant with permanent memory running on infrastructure the user owns. Brand, site, product, backend and conversational design.
Live at coherence.com: brand identity generated with AI agents we directed, the motion system and website for the group behind Sonora, AboutAI, WakeUp and MudShadows. Agent work remains in build.
Live at domainbooth.com for DomainBooth, a premium domain brokerage: a new animated mark, every page redesigned and rebuilt with designed mobile layouts, and a custom CMS for its sales record, featured names, press and enquiries.
We start with your criteria and a sample of past applications, and audit before anything goes live.
30-minute technical call · no deck