In the first recruitment stage AI most often does three things: extracts data from CVs, compares them with the job requirements and ranks candidates by fit. According to OLX Praca documentation the tool analyzes a candidate's CV and compares it with the job offer, and according to Apps365's description of AI screening systems a similar mechanism includes CV parsing, matching with the posting and candidate scoring.
This approach makes sense especially where a recruiter reviews a large number of applications and needs a quick preselection. In practice AI does not replace the decision; it shortens the time to reach a shortlist and standardizes evaluation based on the same criteria.
According to Apps365 the typical process starts with parsing the CV, converting a PDF, Word or other format into structured data: experience, education, certifications, skills and work history. The system then compares these data with the requirements from the posting, taking into account mandatory competencies, years of experience, education level, and sometimes softer signals such as career path or industry.
In OLX Praca's description an important aspect is tailoring the document to a specific offer: the tool highlights areas needing improvement and suggests how to better tune the CV to the posting. From an HR perspective this means the same mechanism can support both evaluating incoming applications and analyzing why a candidate scores poorly against a role's requirements.
Practical materials from Condu and Lessie repeat one condition: criteria must be defined before running the AI. First list must-have and nice-to-have requirements, and only then set scoring so the system filters according to explicit rules rather than opaque similarity to historical profiles.
Recruitment must account for EU regulations. According to Forsal Annex III to the AI Act covers systems intended for recruitment or selection of natural persons, including placing targeted ads, analyzing and filtering applications and assessing candidates. That means a CV screening tool is not a mere operational add-on but a high-risk technology when used for selecting people.
The main practical risks are automatic rejection without human oversight and perpetuating biases from historical data into new decisions. Therefore, according to Condu the system should act as a filter based on pre-defined criteria, and the outcome must be verified by a human before any rejection decision.
Lub System helps B2B companies implement AI, automation and IT solutions end-to-end - from strategy to deployment. See our services or get in touch to discuss your case.