As soon as AI enters the interview process, the same question comes up: how far can it go without replacing human judgement? The answer depends less on the technology than on the task you give it.
What AI does well
Transcribing a conversation, turning it into a structured summary, preparing targeted questions from a job description and a CV: these are mechanical, time-consuming tasks where AI never tires and never varies from one interview to the next. That is exactly what Lia does during a call or a WhatsApp exchange: she runs the conversation from a script, then returns the transcript and summary — the same script produces comparable output for every candidate on a given role.
Where the line should be drawn
What AI should not do is decide on its own who gets rejected. That is as much a reliability question as a legal one: Article 22 of the GDPR prohibits decisions producing legal or similarly significant effects on a person from being based solely on automated processing. In Fabskill, every candidate can ask to be called back by a human at any time, and no application is rejected automatically.
- Score and signals — AI rates and documents, it does not decide.
- Transparency — the candidate knows from the start that they are talking to an automated assistant.
- Human recourse — a callback from a recruiter is always available.
The right reflex
Faced with a new use of AI in recruitment, the question to ask is not only "does it work technically?" but "who keeps the decision, in the end?". As soon as the answer becomes blurry, that is the signal to put a human back in the loop.