How AI Scoring Works

AI scores answers. People make every hiring decision.

• Never decides on its own: springhire's AI never hires, rejects or advances a candidate. It gives the employer a score with its reasons, and the employer decides what happens next.
• The one automatic step: a qualifying question the employer writes, such as "Do you have the right to work in the UAE?", with the answer they require. It is a fixed rule set by the employer, not AI, and the employer can still advance anyone it stopped.
• Open and editable: every AI score shows its reasons and evidence, and the employer can change it.

• Video answers: we turn the candidate's speech into text with Whisper, then a language model scores that text from 0 to 100 against the criteria the employer wrote for the question, with a short reason and which criteria were met. Without criteria, it judges how relevant, specific and well structured the answer is. The employer can watch the video and read the full transcript.
• Multiple-choice questions and skills tests: marked against the answer key. No AI is involved.
• Short written answers and qualifying questions: checked against the answer the employer set. No AI is involved.
• Resumes: the AI checks the resume against each criterion the employer chose and notes whether it is met, with the evidence it found.
• Personality questionnaires: shown as a profile, never as a pass or fail score.
• Overall score: a weighted average of these parts, using weights the employer chooses. Each part is shown on the report, so the number is never a black box. See a sample report.

The scoring model receives only the question, the employer's criteria and the words of the answer. It never receives the candidate's name, face, photo or video.

• Faces and emotions: no analysis of faces, facial expressions, emotions or body language.
• Voice: no judgement of voice, tone, accent or speaking speed. A strong accent can make speech harder to transcribe, so employers can read the transcript, watch the video and correct a score.
• Protected characteristics: the AI is not asked about age, gender, race, religion, disability or any other protected characteristic, and is told to judge the substance of an answer, not filler words.
• Integrity signals: leaving the assessment tab and webcam snapshots are shown to the employer as notes. They never change a score.

• The full picture first: employers see every answer, transcript and recording next to the score and its reasons before they act.
• Scores can be changed: employers can edit any AI score, and the report shows that it was changed.
• People act, not the system: moving a candidate forward, rejecting them or sending a rejection message are actions a person at the employer takes. springhire does not take them on its own.
• Ranking is a starting point: candidates are ranked by score to save reading time, but the ranking is not a decision.

• The same test for everyone: every candidate for a job answers the same questions and is scored against the same written criteria.
• Silence is never scored: recordings with no speech are marked "No audio captured" and the employer is told, so a muted microphone is never read as a poor answer.
• A wrong language is explained: an answer in a different language from the job's scores 0 with a note saying why, and the transcript stays for the employer to read and re-score.
• Everything can be checked: scores and transcripts stay on the report, so any score can be reviewed later.

• You are told first: before you start, the assessment says that parts of it are scored with AI and that the employer makes the final decision.
• Ask for a human review: you can ask the employer that invited you to have a person review your result, explain it, or delete your data. You can also write to support@springhire.io and we will pass your request on.
• Prefer no AI? Tell the employer before you start, so they can offer you another way to apply.
• Recordings are deleted: video and audio are deleted after the employer's plan period, from 14 days to one year. Scores and transcripts stay with the employer until they delete them. Our Privacy Policy has the details.

Speech-to-text and scoring run on Microsoft Azure OpenAI. Under Microsoft's data terms, the answers we send are not used to train their models. Our Privacy Policy explains what data we keep and for how long.

Using AI in hiring comes with duties for the employer in some places. In short, and not as legal advice:

• European Union: AI used to evaluate candidates is classed as high-risk under the EU AI Act, with employer duties such as human oversight and informing candidates applying from 2 December 2027. Under the GDPR, candidates can ask for human review of decisions about them.
• United Kingdom: the ICO expects meaningful human involvement in recruitment decisions and a way for candidates to challenge them.
• New York City: Local Law 144 asks employers to commission an independent bias audit and to notify candidates at least 10 business days before using an automated tool. Write to us and we will help your auditor with the score data they need.
• Illinois: the Artificial Intelligence Video Interview Act asks employers to tell applicants that AI will analyze their video interview, explain how it works (this page does), get their consent first, and delete videos within 30 days when asked. Since 1 January 2026, employers must also give notice when AI is used in employment decisions.
• California: anti-discrimination rules on automated decision systems apply to employers, including keeping related records for four years.
• Colorado: rules on automated decisions in employment are due to apply from 1 January 2027.