AI Transparency

Classification under the EU AI Act

AI systems used to evaluate applicants are classified as high-risk AI (Annex III No. 4). We therefore rely on transparency, human oversight, logging and traceable scores.

Where AI is used

• Simulated prospect: the prospects (Mag. Dana Gruber in German, Sarah Jenkins in English) are AI. We point this out before every call (Art. 50).
• Transcription: your speech is converted to text live. The AI picks one excerpt (max. 25 sec.) where you were particularly convincing as a voice sample; the rest of the recording is not stored.
• Scoring: a language model scores the message and transcript against fixed criteria.

What the AI does not do

• No emotion recognition, no voice, accent or biometric analysis (prohibited under Art. 5).
• No consideration of age, gender, origin, name or other protected characteristics.
• No automatic hiring or rejection decision.

Criteria

The same for everyone in a language: the same scenario, the same scale from 1 to 100. First outreach: personalisation, value proposition, call to action, tone. Live call: cold calling & opening, objection handling, discovery & questioning, expression & persuasiveness, coachability. Every score must be backed by a verbatim, timestamped quote, otherwise it is discarded.

Your consent and rights

Before your first assessment you confirm that you have read this information. You can withdraw consent to storing the voice sample (Art. 6(1)(a) GDPR) at any time in the assessment area — your assessment data is then deleted, or removed after 12 months of inactivity at the latest. Name and education stay hidden from companies until you accept an interview request; email and LinkedIn are never shared.

Human oversight

You see your full assessment before publishing and can repeat it or publish nothing. Companies see the same evidence and decide themselves. You can request a human review at datenschutz@smartscale.hr.

Limits

AI scores can be wrong, for example with poor audio quality or an unusual course of conversation. They reflect a 10-minute snapshot, not overall suitability.