AI interviews that find the right candidate, and prove it's really them

The AI interview is a structured conversation led by a virtual interviewer, via chat or voice call, covering the topics your company defines. It replaces the screening interview, that first 10-, 20- or 30-minute conversation your recruiting team repeats dozens of times per role. The AI introduces itself, explores each topic, asks for concrete examples when an answer is shallow, and wraps up once it has covered everything or reached the configured time limit. Everything is recorded and transcribed, and the recruiter can review it whenever they want. It does not replace the interview with the hiring manager or the hiring decision.

No. The AI scores each criterion configured by the company and generates a summary and written feedback, and the recruiter sets the final score and records their notes. Automatic review helps people decide faster and more consistently; it is not a filter that runs on its own. The company can set up knockout topics, but they work as objective requirements you define (availability, certification, salary expectations), not as a subjective judgment by the AI. This separation matters for data protection compliance, such as Article 20 of Brazil's LGPD: decisions that affect the candidate keep human review.

The candidate receives the invitation, opens the assessment and is told right at the start whether the interview will be by text or by voice. In chat mode, they read the questions and type their answers. In voice mode, they speak into the microphone and hear the interviewer, with automatic detection of when they start and stop talking, no button needed. They can interrupt the AI to add something, see an audio indicator confirming they are being heard, and use a mute button for quick pauses. If the connection drops, the session reconnects without losing the history. Voice mode requires headphones and an up-to-date Chrome, Edge or Firefox.

You create an AI Conversational Interview question in the library, choose the format (chat or voice) and the interviewer's avatar, and build the topics: each one has a name, a guiding question, evaluation tips and follow-up questions that tell the AI where to dig deeper. Then you define the evaluation criteria, each with a description and a weight from 1 to 5, and you can mark bonus criteria that reward exceptional answers without penalizing candidates who don't reach them. The platform automatically generates criteria from the topics as a starting point. The candidate never sees the criteria. An assessment can include up to three conversational interviews, depending on your plan, and can combine the interview with technical and behavioral tests.

The AI interview reduces one kind of bias and calls for care with another. It reduces bias because every candidate answers the same topics, is evaluated on the same criteria with the same weights, and the score for each criterion is recorded and auditable, something a phone screen never offers. It calls for care because language models can react to speaking style, accent or vocabulary. That's why the evaluation is based on the transcript, not on voice or image, the criteria are written by the company instead of inferred by the AI, and the final score goes through a recruiter. We recommend periodically reviewing the score distribution across groups, as you would with any hiring stage.

It depends on the mode. In chat, the risk exists: someone could paste the question into another tool and bring back the answer. For this mode, it's worth enabling tab monitoring and copy-paste logging in the assessment's Integrity tab, which show the recruiter whether the candidate left the screen or pasted text. In voice mode, cheating is much harder: the conversation happens in real time, the AI follows up on the spot, and anyone reading a ready-made answer stalls when the interviewer asks for a specific example. The recruiter has access to the recording and the transcript and can notice pauses and changes in pace. For high-volume screening, voice mode is the most robust choice against this kind of fraud.

Yes. The AI interview runs inside a Coodesh assessment, and the assessment's integrity features apply to it: tab-exit monitoring, full-screen mode, second-screen detection, copy-paste logging, browsing recording, webcam with periodic snapshots and photo ID at the start. In voice mode, the audio recording itself is the main record, so in most cases photo ID and tab monitoring are enough. Our recommendation is to be proportional: turn on what makes sense for the role's risk level, and let the candidate know. Coodesh's integrity page details each feature.

The experience was designed to feel like a conversation, not a form, and that makes a difference in completion rates. The AI introduces itself, signals that it is listening, handles interruptions and closes politely. The candidate knows from the start that they are talking to an AI and that a recruiter will review the results. What usually drives drop-off isn't the AI itself, it's an interview that is too long or lacks context: keep it to three to five topics, and explain in the invitation why the company uses this format and what happens next. Transparency up front reduces abandonment more than any technical tweak.

The interview generates a transcript, an audio recording in voice mode, and scores for each criterion. This data is linked to the assessment and hiring process in which it was collected, with access controlled by your company. Before starting, the candidate is informed that the conversation is conducted by AI and recorded. The legal basis and purpose are defined by the hiring company, which acts as the data controller, with Coodesh as the processor. Deleting the workspace removes the data within 30 days, according to the privacy policy. For companies hiring in the European Union, note that automated candidate screening is a high-risk use case under the AI Act, which reinforces the importance of keeping human review.

The conversational interview and automatic review consume additional credits based on the number of interviews conducted, on top of your platform plan. The limit of interviews per assessment (up to three) also varies by plan. The economic advantage shows up in time saved: a 25-minute human screen per candidate, for a role with 80 applicants, adds up to more than 30 recruiter hours; the AI conducts them all simultaneously and the recruiter only spends time reviewing the transcripts that matter. To understand credit usage for your scenario, talk to our sales team or check your plan.