We demo Ntrvsta, an AI avatar interviewer for the top of the hiring funnel: dynamic interviews, instant scores, ATS integration and why recruiters stay.
An AI interviewer can run the first-round interview the moment a candidate is ready, score it right away and hand recruiters a recording, transcript and evaluation, which removes the scheduling bottleneck at the top of the hiring funnel. On Episode 6 of Built This Week, Jordan Metzner and Sam Nadler showed Ntrvsta, the AI avatar interviewer they built. Jordan says they have run thousands of interviews through it since launch.
Ntrvsta started as an internal project. Jordan says the team does thousands of interviews a year, and every one is expensive. If the interview leads to a hire, it was a success. If not, some of that time may have been wasted. Interviews take anywhere from thirty minutes to an hour, and interviewers often spend that time with a candidate they already suspect is not a fit, out of respect for the candidate.
The team had already automated resume scoring with Ryz Score, which we covered on Episode 4. That saved recruiters a lot of time, but Jordan says the obvious remaining time sink was interviewing. The idea was simple: use AI and AI avatars to help interview.
He lists what the AI interviewer has that humans don't: unlimited availability, unlimited scalability and a big reduction in bias, with scoring that is evenly fair. It can interview across many subjects, and the conversations are dynamic rather than a fixed set of questions. Interviews can happen in many languages and are easy to share and repeat. The avatar interviewer is named Carolina.
Watch why we built an AI interviewer in Episode 6 (starts at 3:03) · Episode 6 notes and transcript
Jordan says candidates have given a lot of positive feedback. They know it is an avatar, and while it can feel different or even awkward at first, they adapt quickly and find that she is patient, calm and there to help. He also credits transparency: after the interview, each candidate gets an email with their score, which he says builds buy-in from the candidates themselves.
His larger argument is about the funnel. To find the best people you need to open up the top of the funnel, which means more candidates and more interviews.
And so, AI is kind of the only way that you can truly kinda scale your organization at the top of the funnel through interview process.
— Jordan Metzner, Episode 6
For Sam, speed is the first win. A candidate can go from a resume score to taking an interview within minutes, instead of waiting days or even weeks for a recruiter's calendar and the candidate's calendar to line up. Scores arrive immediately on several criteria, so strong candidates can be prioritized faster than in a normal process.
Jordan asked him directly whether it has replaced recruiters. Sam's answer:
So there's still a human element to our process. We still do human based interviews within our funnel, so it hasn't completely replaced the recruiters.
— Sam Nadler, Episode 6
Recruiters still manage open roles and do outbound work to get people engaged. Sam says the focus is on expanding their productivity, not cutting the team. In fact, Jordan says the recruiting team has grown since Ntrvsta was put in place, because each recruiter can be more efficient.
Jordan demoed a real interview with a dermatologist applying for an RLHF role in dermatology. She scored 89.5 overall, with a 90 on technical skills and 85 on soft skills. The review page starts with the candidate's background and interests, a headshot, and highlights and lowlights from the interview.
Reviewers can watch the whole interview or jump straight to any chapter. There are subtitles in multiple languages, adjustable playback speed, and every question listed with the answer and an evaluation. The full transcript can be copied. Some interviews include a whiteboard section where the candidate sketches something, in this case a workflow diagram, and the drawing is analyzed too. Sam noticed the avatar asks follow-up questions based on the candidate's previous answers, and says the latency is very reasonable.
Sam also asked how the avatar stays on topic for dermatology, engineering or legal interviews. Jordan says the team did a lot of prompt engineering to keep the interviewer friendly and on track, and tried hard to make it say something inappropriate, with really good results so far.
Most of the team never opens the Ntrvsta interface. They use it through a direct integration with their applicant tracking system, which for the team is Lever. Jordan says it also integrates with Greenhouse, Ashby, Workday and Salesforce, so it acts as a secondary view while recruiters keep moving candidates through their usual flow.
To create an assessment, you upload a job description, pull one from the ATS or CRM, or enter it manually with topics to cover, specific questions and instructions. You can turn on whiteboard mode for interviews that need it. Sam points out you can edit the generated assessment before sending it, or ship it as is. The product also has user management, dashboards and analytics.
Sam noted that this one was not vibe coded in a few days by one or two people. Jordan says the AI toolset improved significantly while they were building it, and they keep upgrading as new technology comes out. Ntrvsta uses ChatGPT and other AI tools for scoring, translation and feedback. On top of that sit text to speech, speech to text, avatar generation, video streaming and video editing. The goal is an end-to-end experience that feels as close to a real meeting and a real interview as today's technology allows. More on the build is on our Ntrvsta build page.
Jordan thinks it is only a matter of time:
I think we're gonna see a lot more AI interview technology implemented across pretty much every major company. And I think mostly because of the fact that it actually really levels the playing field.
— Jordan Metzner, Episode 6
He compares it with Amazon, where he had to be trained to give interviews, score each one and then sit in a debrief with four or five people. He expects AI interviewing to remove a lot of biases so the best candidates rise to the top, not just whoever made someone feel a certain way.
Sam says the team was a little apprehensive at first and now loves it. The usual fear with AI is that it will take jobs, but in the last thirty days, he says, they increased their number of recruiters by 25%. For more on how AI is changing hiring, see our AI in recruiting and HR page and Episode 9, where Sam built an applicant tracking system with Claude Code.
The episode includes the full interview review demo, a clip of the avatar interview itself and how assessments are created. Watch Episode 6 of Built This Week, where the full transcript is also on the page.
