AI Interview
An AI interview is a pre-recorded or live automated conversation where software evaluates a candidate's answers against predefined criteria. In Switzerland, you must disclose its use and ensure it doesn't discriminate. The most common mistake is deploying AI screening without testing it for bias, which can violate employment discrimination law and damage your employer brand.
An AI interview uses software to conduct initial screening conversations with candidates. The system records responses, analyzes tone and word choice, and scores applicants automatically. It replaces human conversations in the first selection round.
SMEs in Switzerland increasingly use AI interviews to handle volume when you receive 200+ applications per role. However, Swiss employment law and the revised Data Protection Act (revDSG) set clear boundaries on how you can use them.
When AI interviews come up
You typically consider AI interviews when your hiring volume exceeds what your team can screen manually—often 100+ applications for a single Pensum position. Smaller firms sometimes use them for high-turnover roles, like customer service or entry-level positions.
AI interviews work best when you have clear, objective scoring criteria. They fail when your role requires judgment calls, cultural fit assessment, or nuanced communication skills that software cannot reliably evaluate.
- High-volume recruitment (100+ applications per role)
- Roles with standardized competency requirements
- Multi-stage hiring processes where initial screening is repetitive
- Distributed hiring across multiple cantons or locations
- Budget constraints that make human screening time-intensive
- Need for consistent evaluation across all candidates
Your legal obligations
Swiss law does not ban AI interviews, but you must disclose their use to candidates upfront and respect two core requirements. First, under revDSG (revised Federal Data Protection Act), you must explain how the system processes personal data and how long recordings are kept. Second, under the Employment Act (ArG), you cannot use AI screening that produces discriminatory outcomes based on protected characteristics like age, gender, origin, or religion.
Practically, this means you should document your AI tool's validation—evidence that it does not systematically reject candidates from protected groups. If a candidate challenges your decision, you must show that the AI system was fair. Many Swiss cantons and sector-specific collective labour agreements (GAV) have begun publishing guidance on algorithmic fairness in hiring; check your industry association.
- Disclose AI interview use in your job posting and privacy notice
- Keep proof that your AI tool does not discriminate (validation testing)
- Comply with revDSG data retention rules (typically 6–12 months maximum)
- Provide candidates a right to appeal AI decisions and speak to a human
- Check sector GAV terms—some collective agreements restrict AI screening
- Consult a lawyer if you operate across multiple cantons with different labour norms
Most common mistake
The single biggest error is deploying an AI interview tool without first validating it for bias. Vendors rarely disclose whether their software has been tested on diverse candidate populations. Many tools perform poorly on non-native speakers, candidates from non-Western educational backgrounds, or people with accents—all common in Switzerland's multicultural workforce.
If you use an unvalidated AI system and later face a complaint from a rejected candidate, you have no defence. Ask your vendor for a third-party bias audit report before you sign the contract. If they cannot provide one, the reputational and legal risk usually outweighs the time saved in screening.
- Deploying AI without a documented bias audit from the vendor
- Failing to tell candidates they will be evaluated by AI
- Using AI scores as the sole hiring decision without human review
- Not explaining to rejected candidates why the AI rejected them
- Ignoring revDSG consent and data retention requirements
- Assuming an AI tool validated in the US works fairly in Switzerland
Frequently asked questions
- Can I use an AI interview tool if I have only 50 applicants?
- Legally, yes. Practically, no—the setup cost and bias-testing overhead make it uneconomical. AI interviews justify themselves above 150–200 applications per role. Below that, human screening is faster, cheaper, and more reliable for small teams.
- Do I need to tell a candidate they were rejected by AI?
- Yes. revDSG requires you to inform people when an automated decision significantly affects them. If an AI tool rejected their application, you must tell them. Best practice is to offer a brief human review on request—this also protects you legally.
- Which Swiss cantons have special rules for AI hiring?
- No canton has passed specific AI hiring laws yet. However, several sector GAVs (e.g., banking, insurance) have issued voluntary fairness guidelines. Check your industry association. A lawyer familiar with your canton's labour court practice can advise on emerging norms.
General information for Swiss employers, not legal advice. Have a lawyer confirm anything with legal consequences.