Interview intelligence is software that records a job interview, turns it into a written transcript, and gives the hiring team a structured summary of what the candidate showed on each skill. It is part of many technical interview platforms, and it is also sold separately for interviews held on ordinary video calls.
It works in four steps. First, the software records the video call, after the candidate has agreed to the recording. Second, speech recognition turns the speech into text. It also separates who said what, so the transcript shows each speaker's lines. Third, a language model reads the transcript. It compares the candidate's answers with the list of skills that the hiring team set for the role, called the rubric. For each skill, it picks out the parts of the answers that are evidence. Fourth, it writes a summary for the hiring team, with links to the exact moments in the recording.
Say an e-commerce company in Noida is interviewing for a product manager. The interview lasts 45 minutes. The rubric has five skills, including "uses data to decide" and "handles disagreement". The transcript runs to about 6,500 words. For "uses data to decide", the summary quotes the candidate's account of an experiment in which a change to the checkout page raised orders by 4%. For "handles disagreement", it finds only one short answer, and marks this skill as "little evidence". It also notes that the interviewer spoke for 60% of the time, which left the candidate less time to answer.
The software does not decide whether to hire the candidate. The hiring team reads the summary, watches the marked moments if needed, and decides. This arrangement, in which a model drafts and a person decides, is called human in the loop. The final decision and the scores are stored in the applicant tracking system.
It helps in three ways. Interviewers no longer have to type notes while they listen. Every candidate is judged on the same rubric. And the company can see which interviewers ask useful questions and which do not.
The same method is used for sales and support calls, where it is called call transcription and AI summaries.