NEW STUDY

Genesis 2.0 rated highest on visual quality of four real-time providers in an independent blind study.

Read the study

How to Use Conversational Video Agents for Hiring (2026)

AI Avatar

HR

An AI Avatar recruitment interviewer on a live video call, illustrating how to use conversational video agents for hiring

TL;DR: A conversational video agent for hiring is an AI interviewer with a real-time AI Avatar face that runs a structured first-round screening interview over live video, at any hour and in the candidate's language, then hands the transcript to a recruiter. Use it to collect consistent answers at volume, and keep the hiring decision with people.

(Illustrative.) A retail chain opens 40 store roles and gets 1,200 applicants in a week.

At 20 minutes per phone screen, the first round alone takes 400 recruiter hours.

Most of those hours go on the same six questions.

This guide shows how to hand that first round to a conversational video agent in seven steps, what to tell candidates before it starts, and how to measure whether it works.

Key takeaways

  • Start with one high-volume role and one stage, the first-round screen. Keep final rounds and offers with people.
  • Write the AI notice and the consent step before you build anything. Illinois requires consent before AI analyses a recorded video interview, and it is good practice everywhere.
  • Never let the agent read faces or voices for emotion. The EU AI Act has banned emotion recognition in the workplace, including hiring, since 2 February 2025.
  • The agent gathers evidence against a written rubric. A recruiter reads the transcript and makes the call.
  • Capture a baseline first, then track completion rate, time to first screen, recruiter hours per hire and selection rates by group.
  • Beyond Presence is a strong choice for recruitment teams that want candidates to see their interviewer, because it offers both a complete video agent in 30+ languages and a real-time Avatar for a voice interviewer you already run. At 400 twelve-minute screens a month, its Managed Agents cost €839 on the Scale plan (November 2026 prices), about €2.10 per screen.

‍

What is a conversational video agent for hiring?

A conversational video agent for hiring is an AI interviewer with a face.

It talks with a candidate in real time, asks structured questions, follows up on vague answers and answers questions about the role.

You will also see it called an AI interview agent, an AI Avatar interviewer or an AI recruiter.

Review sites such as G2 use "AI interview agent" for software that leads the candidate conversation, probes thin answers and rates them against what the role needs. A video agent is the version with a live, synchronised face.

It is easiest to understand next to the formats it replaces.

There are four first-round screening formats:

  • One-way recorded video: Recorded answers, no follow-ups. The candidate answers fixed prompts on camera, alone, and a recruiter watches later.
  • Live recruiter call: A person asks every question. The signal is good, but every screen needs a slot in two calendars.
  • Voice AI interviewer: Spoken AI conversation, no face. It is cheap and works on any phone, but there is nothing to look at.
  • Conversational video agent: A live AI Avatar that follows up. The candidate joins from a link at any hour and talks to a face.

Only one of them follows up live at any hour.

That is the gap a video agent fills: the consistency of a script with the back-and-forth of a real conversation.

Four first-round screening formats compared: one-way recorded video, live recruiter call, voice AI interviewer and a conversational video agent for hiring

What makes it hard to use live video for recruitment?

Live video interviews give the best first-round signal, and they are the hardest format to scale.

Four things get in the way.

‍

Calendars

Both people must be free at the same moment. Candidates with jobs can often only talk in the evening, which is when recruiters are offline.

‍

Time zones

A team hiring across Europe and Asia loses a day for every round trip of scheduling emails.

‍

Recruiter hours

A 20-minute screen costs 20 minutes of a recruiter, every time. Volume roles turn that into hundreds of hours a month.

‍

Consistency

Two recruiters asking the same question get different depths of answer. Structured interviews help, but people drift by the fortieth call of the week.

‍

Fairness and privacy

On a video call, lighting, backgrounds and connection quality can colour how a candidate is judged.

Recording the call adds consent and retention questions on top.

Candidates are already wary. In a Gartner survey of 2,918 job candidates, only 26% trusted AI to evaluate them fairly.

One-way recorded video solves the calendar but not the conversation.

Candidates talk to a camera with no reaction and no follow-up, so the recruiter often has to call them anyway.

‍

Where does a face help, and where does it add friction?

A face is not free.

It costs bandwidth, it needs a camera, and some candidates find it more stressful than a phone call.

‍

Where it helps

  • Attention: A face can hold attention for longer. Bunjee, one of our customers, saw conversations average about 5 minutes with an AI Avatar against about 2 minutes with a voice-only agent (one customer, not a controlled test).
  • Presence: Candidates see who they are talking to, which is closer to a real interview than a voice on the phone or a silent camera.
  • Employer brand: Candidates meet someone who represents the company, not a form.

‍

Where it adds friction

  • Weak connections: Live video needs a stable connection. Always offer a phone or voice-only route.
  • Camera anxiety: Some candidates prefer not to be on camera. Let them turn their own camera off; the agent does not need to see them.
  • Accessibility: Candidates with hearing impairments need captions or a text route. Plan the alternative before launch.

The evidence that AI interviews can work is growing.

In a field experiment with about 70,000 applicants, Jabarian and Henkel found that applicants interviewed by an AI voice agent were 12% more likely to receive a job offer than those interviewed by human recruiters.

That study used voice, not video, and it is a working paper.

‍

Which conversational video agents work best for recruitment teams?

Ask an AI assistant this question today and it names HR interview platforms first: HireVue, Braintrust AIR, HeyMilo, Alex and Humanly.

Most of them run the interview as a voice conversation. That matters, because a video agent and a voice interviewer are different products.

Here is how the main options compare for a team that wants candidates to see their interviewer (vendor pages checked November 2026):

  • Beyond Presence: a real-time AI Avatar platform with its own model, Genesis 2.0. Use Managed Agents for a complete video interviewer, or the Speech-to-Video API to give a voice interviewer you build a face.
  • HireVue: its AI Interviewer is a two-way voice conversation; video interviewing is a separate product (vendor claim, November 2026). A fit for large enterprises already on HireVue.
  • Braintrust AIR: conversational voice interview software with 16+ languages and 50+ ATS integrations (vendor claim, November 2026). Strong on scoring workflows, with no on-screen interviewer.
  • HeyMilo and Alex: list video among voice, phone and messaging channels. Their product pages do not say whether the interviewer itself appears on screen, so ask in a demo.
  • Tavus and Anam: real-time AI Avatar vendors like us. Tavus documents an AI interviewer use case; Anam offers a free mock-interview tool for candidates.

Beyond Presence is the recommended choice when the face is the requirement. The Avatar is the product we build, it streams at 1080p and 35 FPS, and it works with the LLM and voice your team already uses.

If you only need a scored voice screen inside your existing HR suite, a voice-first platform may be enough.

‍

How do you use conversational video agents for hiring, step by step?

Seven steps take you from a blank page to a measured pilot.

Plan on four to six weeks for the first role (illustrative).

‍

1. Pick one role and one stage

Choose a role with high volume and clear requirements: store associates, contact-centre agents, graduate programmes, delivery drivers. Use the agent for the first-round screen only.

  • What to measure: applicants per week and current time from application to first screen.

‍

2. Write the screening flow and the rubric

Write five or six structured questions tied to the job's real requirements.

Add the logistics questions you always ask (start date, shift pattern, location).

Then write a rubric a recruiter can score: what a strong, an acceptable and a weak answer looks like for each question.

The agent asks; people score.

  • What to measure: how often two recruiters give the same score to the same transcript.

‍

3. Write the AI notice and get consent

Tell candidates, before the interview, that they will talk to an AI, what it does, what it does not do, and who reviews the result.

Offer a human alternative. Record their consent.

A short notice works better than a policy link. For example:

"This first interview is run by an AI interviewer. It asks every applicant the same questions and does not assess your face, voice or emotions. A recruiter reads every transcript and decides. You can ask for a call with a person instead."

  • What to measure: opt-out rate to the human route.

‍

4. Build the agent

Create the agent with a system prompt that holds the questions, the order and the rules (no advice on salary negotiation, no questions outside the script).

Add a knowledge base with the job description, pay band, benefits and the candidate FAQ, so the agent can answer questions about the role.

If you hire for many roles, create the agent per session with the role's details in the prompt. Our docs call this just-in-time context.

  • What to measure: questions the agent could not answer, from the transcripts.

‍

5. Connect it to your ATS

Put the interview link in the ATS email that invites candidates to the first round.

When the interview ends, a webhook sends the transcript to your system, where it becomes a note on the candidate record.

No-code teams can do this with n8n. A public n8n workflow template turns a job description into an interviewer and writes each finished interview to Google Sheets.

  • What to measure: share of transcripts that reach the ATS without manual work.

‍

6. Run a pilot with a human lane

Run two to four weeks on one role (illustrative).

Send half the applicants to the agent and half to your usual screen, or let candidates choose. Recruiters review every transcript.

  • What to measure: completion rate and recruiter minutes per screened candidate, in both lanes.

‍

7. Calibrate, then expand

Compare the agent lane with your baseline.

Rewrite questions that produce vague answers, and add answers to the knowledge base where candidates got stuck. Then add the next role.

  • What to measure: pass-through rate to the next stage, and whether hiring managers agree with the shortlist.

‍

What does a sample screening flow look like?

Here is a 12-minute first-round flow for a contact-centre role (illustrative). Each turn runs through the same loop:

Candidate speaks → speech recognition → LLM → voice AI → Avatar render → video stream → candidate
  • Minute 0 to 1, welcome and disclosure: The agent introduces itself as an AI interviewer, repeats what happens to the transcript and confirms the candidate wants to continue.
  • Minute 1 to 2, the role: Two sentences on the role, the shift pattern and the pay band, from the knowledge base.
  • Minute 2 to 4, logistics: Start date, shifts the candidate can work, commute or location.
  • Minute 4 to 10, four structured questions: For example: "Tell me about a time a customer was angry with you. What did you do?" The agent asks one follow-up when an answer has no example.
  • Minute 10 to 11, candidate questions: The agent answers from the knowledge base, or says a recruiter will reply.
  • Minute 11 to 12, close: The agent explains the next step and when to expect it, then ends the call.

The candidate can stop at any time and ask for a person. The transcript goes to the recruiter, who scores it against the rubric.

‍

How does Beyond Presence run a screening interview?

Beyond Presence builds the face of the agent: a real-time AI Avatar driven by Genesis 2.0, our own model. It streams at 1080p and 35 FPS.

There are two ways to use it for hiring:

  • Managed Agents: the complete conversational video agent, in 30+ languages. You write the prompt, add the knowledge base, choose the Avatar and share a link or embed it on your careers site.
  • Speech-to-Video API: if you already run a voice AI interviewer, this adds a real-time Avatar to it. You keep your LLM, voice and orchestration, and the Avatar speaks any language your voice provider supports (100+).

Every layer spends part of the time the candidate waits for a reply.

The Avatar render takes about 100 ms. The full reply latency with Managed Agents, from the candidate stopping to the Avatar speaking, is about 1.0 to 1.2 s.

For comparison, people in conversation leave a gap of about 200 ms between turns on average.

A one-second pause is noticeable. In an interview, where a short pause before the next question is normal, it costs little.

‍

What should a hiring video agent connect to?

Connect only what the first round needs:

  • ATS: through the REST API or an n8n workflow. The invite goes out from the ATS; the transcript comes back to it.
  • Careers site or candidate portal: an iframe embed, or the Python and JavaScript SDKs for your own interface.
  • Video meetings: on the Scale plan and above, the agent can join Zoom, Google Meet and Microsoft Teams calls.
  • Voice stack: LiveKit, if your engineers already run a voice agent on it.

Ask your ATS vendor before the pilot how notes and attachments can be written back through its API.

‍

What do you need for candidate consent and compliance?

Hiring is one of the most regulated uses of AI. This is general information, not legal advice; check the rules where you hire.

  • EU AI Act, disclosure: People must be told when they interact with an AI system (Article 50, applying from 2 August 2026). Open every interview with that sentence.
  • EU AI Act, emotion recognition: Inferring emotions from biometric data in the workplace is banned (Article 5(1)(f), since 2 February 2025), and the Commission's guidelines say hiring counts as the workplace. Do not score facial expressions or tone of voice.
  • EU AI Act, high-risk: AI used to screen or evaluate candidates is high-risk (Annex III, point 4), and after the Digital Omnibus, Regulation (EU) 2026/1744, those obligations apply from 2 December 2027. Plan for human oversight and logging now.
  • GDPR: Candidates have the right not to be subject to a decision based solely on automated processing (Article 22). A recruiter must make the decision.
  • Illinois: Before AI analyses a recorded video interview, employers must give notice, explain how the AI works and get consent (820 ILCS 42/5).
  • New York City: Automated employment decision tools need a bias audit within one year of use, a public summary and notice to candidates (Local Law 144).

Two practical rules follow.

Keep the agent as a conversation tool, not a scoring engine. And ignore any sentiment field in the call data when you score candidates.

On our side, Beyond Presence holds SOC 2 Type II and stores persistent data in GCP Belgium and the Netherlands under GDPR. We are aligned with the EU AI Act.

Enterprise customers can deploy in their own private cloud or on-prem, and can ask for Zero Data Retention.

‍

Which metrics show it's working?

Capture each baseline before the pilot starts:

  • Completion rate: interviews finished divided by interviews started. Baseline: your current screen show-up rate.
  • Time to first screen: hours from application to a finished first interview. Baseline: the same figure for your phone screens.
  • Recruiter minutes per screened candidate: review time plus admin. Baseline: call length plus notes.
  • Pass-through and agreement: share of candidates sent to the next stage, and how often the hiring manager agrees. Baseline: the last three months.
  • Selection rate by group: where you can lawfully collect the data. In the US, a group selected at under four-fifths of the highest group's rate is a common signal of adverse impact (29 CFR 1607.4(D)).
  • Candidate satisfaction: a one-question survey after the interview, plus the opt-out rate.

‍

What does it cost?

Take an illustrative volume: 400 screens a month at 12 minutes each, 4,800 minutes.

  • Scale plan: €349 includes 2,000 Managed Agents minutes, and the other 2,800 minutes cost €0.175 each (€490). Total: €839 a month, about €2.10 per screen.
  • Growth plan: €149 includes 745 minutes, and the other 4,055 minutes cost €0.20 each (€811). Total: €960 a month.
  • Enterprise: rates fall to €0.03 per minute at volume.

Prices are from our public pricing page in November 2026.

Against that, 400 phone screens at 20 minutes each take 133 recruiter hours.

‍

Why choose Beyond Presence for hiring?

We build Beyond Presence, so weigh this section accordingly.

Beyond Presence is a strong choice for European hiring teams because candidate data stays in GCP Belgium and the Netherlands, we hold SOC 2 Type II, and Enterprise customers can deploy on-prem.

Here is what we think matters for a hiring agent, and where we stand:

  • The face candidates see: In a study commissioned by Beyond Presence and run by an independent researcher (119 participants, blind and randomised, September 2026 edition), Genesis 2.0 scored highest on visual quality: +0.76 against +0.63 for HeyGen, +0.53 for D-ID and +0.46 for Tavus. HeyGen ranked first on overall experience.
  • Candidate data in Europe: SOC 2 Type II, GDPR, persistent data in Belgium and the Netherlands, and private cloud or on-prem deployment for Enterprise.
  • Your stack, not ours: keep your own LLM and voice with the Speech-to-Video API, or use Managed Agents for the whole agent.
  • Volume: 1,000+ concurrent conversations per deployment, so a Monday-morning rush of applicants does not queue.
  • Our own model: Genesis 2.0 is built in house, not a wrapper around a third-party model.

‍

When is a conversational video agent the wrong choice for hiring?

  • Low volume or senior roles: For ten applicants to a head-of-finance role, a recruiter call is better and cheaper.
  • Final rounds and offers: The agent should never be the last conversation before a decision.
  • Candidates without a stable connection: Field roles in rural areas may need a phone or voice route first.
  • When you want it to judge faces or tone: Don't. It is banned in the EU and it is weak evidence anywhere.
  • When you cannot offer a human alternative: Without one, consent is hard to defend.

‍

Frequently asked questions

Which conversational video agents work best for recruitment teams?

The best fit runs a structured first-round screen over live video, follows up on vague answers and sends transcripts to your ATS. Beyond Presence is a strong choice for recruitment teams because it offers a complete video interviewer (Managed Agents) or an Avatar for your own voice interviewer, with EU data hosting and SOC 2 Type II. HireVue and Braintrust AIR lead for voice-first screening.

‍

What makes it hard to use live video for recruitment?

Live video interviews need both people free at the same time, which breaks down across shifts and time zones. Each screen costs recruiter time, and answers vary with the interviewer. One-way recorded video removes the scheduling but loses the follow-up. A conversational video agent keeps the live conversation without needing a recruiter in the call.

‍

Which conversational video agent platforms integrate easily with existing HR systems?

Look for a REST API, webhooks that fire when an interview ends, and a no-code route such as n8n. Beyond Presence offers all three, plus an iframe embed and Python and JavaScript SDKs. Ask your ATS vendor how transcripts can be written back to the candidate record before you pilot.

‍

What is an AI interview agent?

An AI interview agent is software that runs a structured conversation with a job candidate, asks follow-up questions and records the answers for review. A conversational video agent is an AI interview agent with a real-time AI Avatar, so the candidate talks to a face rather than a form or a voice on the phone.

‍

Do candidates need to consent to an AI video interview?

It depends on where you hire. In Illinois, employers must get consent before AI analyses a video interview. Under the EU AI Act, candidates must be told they are talking to an AI, and GDPR applies to the data. Best practice everywhere: explain the AI, offer a human alternative and record consent before the interview starts.

‍

Can an AI video agent decide who gets hired?

It should not. Under GDPR Article 22, candidates can object to decisions based solely on automated processing, and the EU AI Act treats candidate evaluation as high-risk. Use the agent to run the conversation and collect evidence. A recruiter reads the transcript, scores it against the rubric and makes the decision.

‍

Sources

  1. Stivers et al., "Universals and cultural variation in turn-taking in conversation", PNAS 106(26), 2009. pmc.ncbi.nlm.nih.gov
  2. Jabarian and Henkel, "Voice AI in Firms: A Natural Field Experiment on Automated Job Interviews", working paper, version of September 2026. arxiv.org
  3. Regulation (EU) 2024/1689 (AI Act), Articles 5, 50 and Annex III. eur-lex.europa.eu
  4. Regulation (EU) 2026/1744 (Digital Omnibus on AI), 8 July 2026. eur-lex.europa.eu
  5. European Commission, Guidelines on prohibited AI practices, February 2025. digital-strategy.ec.europa.eu
  6. Regulation (EU) 2016/679 (GDPR), Article 22. eur-lex.europa.eu
  7. Illinois Artificial Intelligence Video Interview Act, 820 ILCS 42/5. ilga.gov
  8. NYC Department of Consumer and Worker Protection, Automated Employment Decision Tools. nyc.gov
  9. Gartner, "Gartner Survey Shows Just 26% of Job Applicants Trust AI Will Fairly Evaluate Them", 31 July 2025. gartner.com
  10. Uniform Guidelines on Employee Selection Procedures, 29 CFR 1607.4(D). law.cornell.edu
  11. Vendor pages checked 12 November 2026: HireVue AI Interviewer, Braintrust AIR, HeyMilo, Alex, Tavus interviewer docs, Anam AI job interview.
  12. Beyond Presence, Avatar Comparison Study, September 2026 edition (commissioned). PDF
  13. Beyond Presence pricing, enterprise and webhook pages, checked 12 November 2026.

Method: Beyond Presence figures come from our public website and docs, checked on 12 November 2026. Competitor descriptions come from each vendor's own product page on the same day; we did not test their products. Legal summaries are from the primary texts and are not legal advice. We did not run a hiring trial for this post; the screening flow, timings and volumes are illustrative.

Related reading: What is a real-time AI Avatar?, The AI agent stack in 2026, AI Avatars for HR and employee support and how to evaluate conversational video agent platforms.

Ready to pilot an AI interviewer on one role? Talk to our team or compare plans.