AI recruiting that reads every application before you do.
One job ad brings in more applications than anyone can read with care, and the best candidates sign elsewhere while they wait. We build agents that summarize each application against the criteria you wrote, answer candidates and book interviews in your applicant tracking system. Your recruiters see every application and make every decision.
- Teamtailor, Greenhouse, Lever
- Built for the EU AI Act
- A recruiter decides
AI recruiting and resume screening: what the agent does
AI recruiting means agents that take on the paperwork of hiring. They read each application against the criteria written for the job and summarize it, answer candidates’ practical questions, schedule interviews and keep the applicant tracking system up to date. Screening candidates is high-risk under the EU AI Act, so a recruiter reviews every application and takes every decision, and candidates are told beforehand.
The summary does the slow part of the first read. For each criterion of the job, the recruiter sees what the resume says and the passage it comes from, plus what is missing. Name, photo, age and address are hidden from that view, because they have nothing to do with the criteria and can bias a quick read.
Nothing is rejected by the agent. It never ranks people on a score it invented, never reads a face or a voice, and never decides who gets an interview. It prepares the reading so your recruiters can spend their time on the candidates themselves.
What your recruiters live with today
- We are afraid a tool will reject candidates without anyone knowing why.
- Applications pile up faster than anyone can read them properly.
- Good candidates accept another offer while they wait for an answer.
- Booking one interview takes an endless chain of emails.
- The tracking system holds years of profiles nobody ever searches.
What an agent takes off your recruiters’ desks
Start with the steps that delay every hire. Each of these runs in your applicant tracking system, under the rules set out further down.
01
Resume screening against your criteria
Each application summarized criterion by criterion, with the passage that supports each point and personal details hidden. The recruiter reads every one and decides.
AI agents02
Candidate questions, any hour
An assistant on your careers page answers questions on the role, the process and the dates from your own job description, and says it is an AI.
AI chatbot03
Interview scheduling
Slots proposed from the interviewers’ calendars, invitations sent, reminders and room bookings handled, and the tracking system updated at each step.
Process automation04
Job ads from the manager’s notes
A draft ad written from the hiring manager’s notes, with every wording that could exclude a protected group flagged for your review.
AI integration05
From signed offer to first day
Once a candidate accepts, the HR assistant takes over: documents collected, accounts requested and a first-week checklist sent to the manager.
AI for HR06
Check where you stand
Ten questions, about two minutes, to see whether your company is ready for a first agent and where it would pay off, hiring included.
Take the readiness check
AI candidate screening and the AI Act: task by task
The same regulation treats a scheduling assistant and a screening tool very differently. This is how we classify each task before building it, and who decides at the end.
| Task | What the agent does | Who decides | Status under the EU AI Act |
|---|---|---|---|
| Reading and summarizing applications | Summary against the job criteria, sources shown, personal details hidden | A recruiter, on every application | High-risk (annex III, point 4) |
| Ranking or filtering candidates | Nothing on its own: no automatic rejection | A recruiter, always | High-risk, and GDPR article 22 on automated decisions |
| Interview scheduling | Proposes slots, sends invitations and reminders | The candidate and the interviewer | Can fall outside high risk as a narrow procedural task (article 6(3)), assessment written down |
| Answering candidate questions | Answers from your job description and process | The recruiter, for anything about a decision | Transparency: says it is an AI (article 50) |
| Writing job ads | Drafts from the manager’s notes, risky wording flagged | The hiring manager | Targeting ads at chosen people is high-risk (annex III, point 4) |
| Reading emotions in video interviews | Never built | Not applicable | Prohibited since 2 February 2025 (article 5) |
Illustrative scenario
One recruiter, a busy hiring season, and every application still read by a person
In this scenario, Léa is the only recruiter at a growing industrial services company. Field technician ads bring in far more applications than she can read with care, and candidates wait days for an interview slot.
Fictional names, no figures claimed. What changes for your team depends on your hiring volume and process, which is what the free assessment looks at.
- Léa and the hiring managers wrote the criteria of each job before the agent read a single resume.
- Each application reaches her as a summary per criterion, name, photo, age and address hidden.
- Every application is opened by Léa, and only she moves a candidate to the next stage.
- Candidates are told in the ad and in the acknowledgment email that an AI helps with the first read.
- Interview slots are proposed from the managers’ calendars, and employee representatives were informed before go-live.
The rules an AI recruiting tool has to follow
- High-risk under the AI Act.AI that analyzes, filters or evaluates applications is high-risk (annex III, point 4). Since the Digital Omnibus of July 2026, the deployer’s duties apply from 2 December 2027: trained people overseeing it, relevant input data, logs kept at least six months, workers and candidates informed (article 26). We build to them from the first version.
- A person takes the decision.The GDPR gives everyone the right not to be subject to a decision based solely on automated processing that significantly affects them (article 22). Rejection, shortlist and offer are always a recruiter’s act, and the AI Act adds a right to an explanation of the AI’s role in the decision (article 86).
- Candidates informed, criteria tied to the job.In France, candidates are told beforehand which recruitment methods are used on them, and those methods must be relevant to the job (Labor Code, L.1221-8). No one may be excluded on grounds such as age, origin, sex or place of residence (L.1132-1), which is why those details are hidden from the summary.
- No emotion reading, ever.The AI Act bans systems that infer emotions in the workplace (article 5), and the European Commission’s guidelines apply the ban to candidates during hiring. We build no video or voice analysis of candidates.
Recruiting: your questions
Is AI resume screening legal in Europe?
Yes, under conditions. It is high-risk under the AI Act, with deployer duties from 2 December 2027; the GDPR already forbids leaving a significant decision to the machine alone, and in France candidates must be told about the methods used. We design the tool around those rules from the start.
Can the agent reject a candidate on its own?
No. It summarizes and organizes, and every application reaches a recruiter. Rejections, shortlists and offers are recorded as the recruiter’s decision in your tracking system, with a log of what the agent showed them.
Will candidates know an AI is involved?
Yes. The job ad and the acknowledgment email say that an AI helps with the first read and that a person decides, and the careers page assistant says it is an AI. Candidates can ask how it was used for their application.
Do we need to inform employee representatives?
The AI Act requires employers to inform workers’ representatives before using a high-risk system at the workplace. In France, companies of 50 employees or more also consult the CSE on new technologies (Labor Code, L.2312-8). We prepare the description of the tool for that meeting.
Which applicant tracking systems does it work with?
Teamtailor, Greenhouse, Lever, SmartRecruiters, Workday and most tracking systems with an API, as well as your careers page and the shared mailbox where applications still arrive. Data can stay in the EU, on a model hosted there or on your own servers.
How long before it helps on a real job opening?
A first agent usually goes live in four to eight weeks, tested on past applications whose outcome you already know before it reads a new one. Scope and price are fixed before we start.
Give every candidate a careful first read.
Tell us how many openings you fill in a year and where applications arrive today. We reply within one business day to set up your free 30-minute assessment, including a first look at what the AI Act asks of your process.