Candidate Experience: How AI Video Agents Make Every Applicant Feel Heard

Applying for a job is one of the most personally exposed experiences a person has, and the hiring funnel increasingly responds to that exposure with silence. Somewhere tonight, a candidate is rereading the confirmation email from an application she submitted three weeks ago, wondering whether anyone ever saw it.

Candidate experience encompasses the impressions an applicant forms from the job post to the final decision. For many candidates, the process feels like absence. No recruiting team can be present for every application that lands.

Presence, the feeling that someone is genuinely paying attention and responding to what you actually mean, has become scarce in high-volume hiring. This article focuses on real-time, face-to-face AI conversation as one way to address the moments candidates usually spend waiting.

Candidate experience as a competitive advantage in hiring

More than half of applicants never hear back at all. Employer ghosting hit 53% of job seekers within the past year, a three-year high, up from 38% in 2024, per Criteria Corp survey data in Fortune's ghosting coverage. Recruiting teams can also face application volumes that are hard to answer one by one.

For candidates, the modern application process can feel like machinery, with automated steps and impersonal communication leaving little clarity about what happens after they submit. AI-led formats are common enough that the experience now has to earn trust from the start.

Candidates also want to know when they are talking to a machine. Recruiting leaders should clarify how they use AI in hiring and let candidates opt out of AI-led interview formats, per Gartner's October 2025 guidance, and unclear expectations are a recurring source of candidate frustration.

Against that baseline, responsiveness becomes both a candidate-experience issue and a recruiter-capacity issue.

The business cost of a poor candidate experience

Rejected candidates may have a relationship with the company beyond the job opening. A poor rejection process can sour that relationship, turning an operational recruiting gap into a broader brand problem.

Rejected candidates may also become reviewers and referrers. Their experience can change how applicants talk about the brand, whether they return for another role, and how much weight they give to negative reviews. That is how brand damage can start inside the recruiting process.

The downstream effect shows up at the offer stage: offer acceptance fell to 48% by Q4 2025, down from 85% two years earlier, per Gartner's June 2026 research. Offer-stage damage can begin before a recruiter reads a word, in the stretch handled largely by software.

Real-time video conversation across the hiring journey

Before first human contact, most hiring systems ask candidates to navigate keyword filters, one-way recordings, scheduling emails, and text-first automation. Those formats were never built for presence.

Most make interruption hard, and many leave candidate questions for a later exchange. Format is part of the problem: a one-way recorder feels different from a live exchange.

Tavus is the human computing company, building a Personified Application Layer (PAL) that sees, hears, understands, remembers, and responds in real-time conversation. A PAL gives a candidate something she can talk to and build a relationship with in a live, two-way exchange she can interrupt, with continuity beyond one-off answers.

PAL conversations run on the Conversational Video Interface (CVI), a framework for creating real-time multimodal video interactions with AI. Recruiting teams build on it through the Tavus API.

Screening is the widest bottleneck. A PAL recruiter conducts live, two-way screening conversations around the clock in 42 languages. Knowledge Base retrieval runs in English today.

In talent acquisition software deployments, teams can run 100 candidates through 15-minute screening conversations in a single evening.

Scheduling is also where teams lose time. Tavus can hand off next steps during the conversation by presenting calendar or scheduling components so candidates can select and book slots, while Function Calling and webhooks/callbacks can push structured results into the applicant tracking system (ATS).

Priya, a night-shift nurse applying for a clinical educator role, opens her screening link at 11 PM and asks about parental leave. The PAL recruiter answers from the employer's own benefits documentation through Knowledge Base, Tavus's retrieval-augmented generation (RAG) layer. It retrieves in roughly 30ms, so her answer lands without an awkward pause.

Making AI-led conversations feel human

In high-stakes conversations, timing matters as much as wording. Long silences can feel less like thoughtfulness and more like distance, especially in an interaction such as a job interview.

Inside CVI, the closed loop has four parts. Sparrow-1 governs conversational flow, Raven-1 perceives and fuses the other person's emotional and attentional signals, the large language model (LLM) layer reasons about what to say and do next, and Phoenix-4 renders responsive facial behavior.

Sparrow-1, the conversational flow model, predicts who owns the conversational floor at every moment on raw audio: 55ms median prediction latency, 100% precision and recall across 28 challenging conversational samples, and zero interruptions on benchmark. In a screening call, it reads a candidate's pause as unfinished thought, holds the floor open, then responds when a human listener would.

Raven-1, the multimodal perception system, keeps rolling perceptual context no more than 300ms stale. When a candidate says "I'm comfortable with cold outreach" in a flat voice while glancing away, Raven-1 fuses the flat tone with the averted gaze, catching the mismatch between the words and the delivery. It hands the LLM layer a natural-language description of the mismatch, so the follow-up probes where confidence thinned.

Phoenix-4, the real-time facial behavior engine, generates behavior while it listens: nods, responsive micro-expressions drawn from human conversational training data, and 10+ controllable emotional states at 40fps. When a candidate explains a layoff, the face across the screen registers it in the moment. Phoenix-4 renders facial behavior grounded in what the speaker actually does, with expressions generated for the moment instead of a loop.

Candidate experience use cases across the hiring funnel

A standardized first-round screen can give applicants a structured first-round conversation before keyword filters decide who advances. Users stayed 42% longer and completed 35% more sessions when the interviewer felt real and responsive, per Tavus's mock-interview engagement data with Final Round AI.

Structured interviews can give recruiting teams a way to ask consistent questions and apply the same scoring frame across candidates.

In a PAL-led first round, Objectives can be set with measurable completion criteria, such as gathering candidate information or screening a candidate before the session closes, and Guardrails keep the questions inside compliance scope. Each session produces structured evaluation data and full transcripts, with perception analysis webhook-ready for the ATS, per Tavus's interviewer platform page.

After the interview, Persistent Memory remembers that Priya flagged a night-shift scheduling constraint, so the second conversation opens there instead of restarting.

Final Round AI built its interview practice product on CVI, Knowledge Base, and Tavus's real-time behavioral stack: over 1.2 million practice minutes, 100,000+ active users, and 12-minute average sessions, documented in the Final Round AI story. The team's assessment: "The speed and quality have been critical."

Tracking candidate experience improvements over time

Teams that manage candidate experience as an outcome usually watch a compact set of numbers.

• Candidate NPS (cNPS): willingness to recommend your hiring process, usually split by hired, withdrawn, and rejected candidates.
• Candidate resentment rate: the share unwilling to apply again, refer others, or buy after a bad experience.
• Application completion rate: the share of job ad clicks or started applications that end in a completed application.
• Offer acceptance rate and time to hire: downstream outcomes that show whether the process is helping or hurting close rates.

Whatever the metric, survey non-hired candidates; their feedback is the clearest view into the parts of the process most employers never see.

Building a hiring process candidates trust

Trust starts from a low baseline. Only 26% of job applicants trust AI to evaluate them fairly, in Gartner's candidate survey. Candidates already bring expectations around notice, explanation, consent, auditability, and human oversight.

Introduce the PAL as AI, explain what it evaluates, and keep humans in charge of final decisions. Candidate trust starts with not making the machine pretend to be something else.

Think of Priya at 11 PM, asking the question she'd never risk in a live interview, and getting an answer from a face that was actually attending to her. What she experienced is presence, the thing the candidate rereading her confirmation email never got.

Hiring has always come down to whether people feel seen by the companies that ask so much of them. That has always been true; the difference now is that presence does not have to be reserved for the few candidates a recruiter can call.

See it for yourself. Book a demo.

Frequently asked questions

What is candidate experience in recruiting?

Candidate experience covers every impression a job seeker forms of an employer, from the posting to the accept-or-reject call. It breaks down most often after someone hits submit, when weeks pass with no word either way.

How do PALs improve candidate experience?

They can move hiring's silent stretches into live, face-to-face conversation for high-volume screening. A PAL screens applicants on demand, answers policy questions from the employer's documentation, and remembers returning candidates. Candidates can get an on-demand first response instead of relying only on asynchronous follow-up.

Are AI-led interviews fair to candidates?

They are more defensible when structured, disclosed, and paired with human final decision-making authority. Clear explanations, consistent questions, and human oversight matter more than pretending the AI is not there.

Do AI video agents replace human recruiters?

No. A PAL can take over what a keyword filter or an unanswered inbox was doing badly; the approach replaces bad machines. Recruiters can redirect those workflows toward relationship building and final-round judgment.

How is candidate experience measured?

Most teams combine candidate NPS, resentment rate, application completion, offer acceptance, and time to hire, from post-stage surveys and ATS funnel data. The strongest programs survey rejected candidates too, within 24-48 hours of each stage.