Automated screening interview that talks to every applicant

It is the first-round screen, conducted by software instead of a recruiter. The candidate has an actual conversation by voice or video: the agent asks the structured questions your team defined for the role, listens, and asks a follow-up when an answer is thin or ambiguous. Each response is scored against your rubric with the transcript attached as the evidence for that score.

The word automated is doing less work than it sounds like. What is automated is the conducting and the scoring, not the deciding. The output is a ranked list a recruiter reads, with every score traceable to something the candidate said, and a person chooses who moves forward. Nobody is rejected by the system.

Resume screening scores a document. An automated screening interview scores a conversation. That distinction has become sharper over the last two years, because candidates now write resumes with the same class of model that reads them, so keyword matching increasingly measures prompt quality rather than the person behind it.

The practical consequence is which direction each one points. A filter tells you who to exclude, and it excludes strong candidates with unconventional backgrounds first, because those are exactly the resumes that do not match a pattern. An interview tells you who to include, based on what someone can explain about work they have actually done. Filtering narrowly on genuine eligibility, such as work authorization or a required license, and then interviewing everyone who clears that bar is the combination that works.

It depends entirely on the format, and this is where most disappointment in the category comes from. A one-way video recorder removes the scheduling but leaves you watching every submission, so 40 applicants at eight minutes each is still over five hours of somebody's week. The work moved from a calendar to a queue.

A screening interview that scores as it goes removes the reviewing too. You open a ranked shortlist, read the transcripts for the candidates near your cut line, and spend the recovered hours on the parts of recruiting that need a person, such as closing a candidate who has another offer. Our comparison of one-way video interview software sets out where recording genuinely wins.

Structured beats unstructured on consistency, which is the main fairness lever available at this stage. A human screen drifts: the fourth call of the afternoon is not the first, questions vary by interviewer, and rapport with a candidate who reminds you of yourself is hard to switch off. Software asks the same questions in the same order every time, which is trivial for it and genuinely hard for people.

That is not the whole story, and the honest version matters. Consistency does not make a rubric job related, and an automated screen can encode a biased criterion perfectly consistently. What keeps it defensible is the same set of practices as any structured process: job-related criteria, anchored scoring levels, disclosure and consent, an independent bias audit where the law requires one, and a human making the decision. Our AI hiring compliance page covers the employer duties, and our guide to writing a rubric that survives a bias audit covers the criteria themselves.

Ten to twenty minutes for the candidate, covering five to seven questions. For the employer the elapsed time is the more useful number: interviews start the day someone applies rather than the day a recruiter has a free slot, so a pipeline that took two weeks to screen tends to clear in two or three days.

Setup is the part worth budgeting properly. Defining the role, writing the questions and anchoring the rubric takes a couple of hours per role, and it is the step that decides whether any of the output is useful. Teams that rush it get consistent scores against criteria they never really agreed on. Our page on candidate interview scoring covers how to build the rubric.