AI Recruiting That Listens: Conversational Screening + Email Verification

When job applications become a spam minefield, you build something that separates signal from noise.

The Problem

Hiring is broken.

Small teams drown in applications. Most are spam. The rest are half-baked. Finding the 2% worth talking to means sorting through 98% noise.

What we needed:

• Filter spam automatically (fake emails, test submissions, bot traffic)
• Get real information from real candidates
• Stop wasting hiring managers' time
• Make it feel human, not like filling out a form

What we had:

• Standard forms that anyone could spam
• No verification (test@test.com got through)
• Manual screening eating hours per week
• Rigid forms that felt robotic

Companies that hire at scale face this 10x worse. Hundreds of applications per week. Most are garbage. The good ones get lost.

The Solution

We built two systems that work together:

1. Conversational AI Recruiter

A LangChain-powered agent that has real conversations with candidates. Ask questions, listen to responses, adapt based on what they say. Gather the information we need without making them fill out rigid forms.

It works like talking to a real recruiter:

• Candidate: "I'm interested in sales roles"
• Agent: "Tell me about your experience with customer-facing work"
• Candidate: "I worked in restaurants for 5 years, managed a team"
• Agent: "That's exactly what we look for. Service industry experience translates perfectly to tech sales. What motivates you to make the switch?"

Natural. Conversational. Actually listens to their answers.

What it gathers:

• Name and contact info
• Work experience
• Skills and expertise
• Career motivations
• Why they want to work here
• Availability

What it does with the information:

• Submits application on their behalf
• Sends them verification email
• Sends hiring managers AI assessment
• Flags spam vs real candidates

2. Email Verification Pipeline

Every submission (form or recruiter chat) requires email verification before hitting hiring managers.

The flow:

1. Candidate submits (form or conversation)
2. System sends verification email to candidate
3. Hiring managers get "PENDING" notification immediately
4. Notification includes AI assessment: "This looks real/fake because..."
5. Candidate clicks verify link
6. Candidate gets confirmation: "Thanks, we'll be in touch"
7. Hiring managers get "VERIFIED" notification

What gets filtered:

• test@test.com submissions
• Disposable email addresses
• Bot submissions
• Duplicate attempts
• Spam patterns

Hiring managers see every submission but know which ones are verified vs pending. Spam gets flagged automatically.

The Technical Architecture

Conversational Layer (LangChain)

Model: Meta-Llama-3.1-70B-Instruct-Turbo via Together AI

Design principles:

• Maximum 5 questions before contact collection
• One question per response (no interrogation mode)
• Adaptive based on candidate responses
• Intent recognition (sales vs tech vs custom role)
• Fast-track option for candidates who want to skip ahead

Routing logic:

• Sales interest → Route to BDR/SDR application
• Tech interest (AI/ML, Engineering, Product, Design) → Route to tech application
• Other (Marketing, Ops, Creative) → Route to pitch-your-role application

The agent determines which application to submit based on the conversation.

Verification Layer (Database + Email)

Database: PostgreSQL on Vercel (Neon)

Tables:

• lead_verifications - Stores tokens, tracks verification status
• submission_attempts - Tracks all submission attempts by session/IP
• session_blocks - Manages temporary blocks for spam patterns

Email system: Resend API

Verification tokens:

• Crypto-secure random generation
• 7-day expiration
• One-time use
• Tied to email + form type

Admin notifications:

• PENDING: Sent immediately when submitted (shows "⏳ NOT VERIFIED YET")
• VERIFIED: Sent when candidate clicks verification link (shows "✅ VERIFIED")

Both include AI assessment explaining if the candidate seems legitimate and why.

AI Assessment Layer (Together AI)

Every submission gets analyzed:

What it evaluates:

• Email domain legitimacy
• Message coherence and quality
• Experience alignment with role
• Red flags (generic responses, suspicious patterns)
• Green flags (specific details, genuine interest)

What hiring managers see:

• "I think this candidate is legitimate because they provided specific company details and clear role alignment"
• "This looks suspicious. Generic email, copy-paste motivation, likely spam"
• "Strong candidate. Service industry background matches sales role perfectly, motivation is authentic"

The AI explains its reasoning. Hiring managers make the final call.

What This Solves

For Us (Virgent AI)

Before:

• Spam applications reaching hiring managers
• No way to tell real vs fake before opening
• Manual verification eating time
• Good candidates mixed with garbage

After:

• Zero spam reaching hiring managers (verified only)
• AI flags quality before we open it
• Automated verification pipeline
• Real candidates get conversational experience

Time savings: Hours per week not sorting spam.

For Companies At Scale

Organizations that hire at scale face this 100x worse.

The bottleneck:

• Recruiting teams spend 60-80% of time on unqualified applicants
• ATS systems do not filter intelligently
• Resume screening is manual and time-intensive
• Good candidates ghost during rigid application processes

What this system enables:

• Conversational application experience (reduces drop-off)
• Automatic spam filtering (verified emails only)
• AI pre-screening with reasoning (not black box)
• Real-time admin notifications (PENDING + VERIFIED)
• Hiring managers focus on verified, AI-assessed candidates only

The Conversational Advantage

Standard recruiting forms:

• Rigid fields
• High drop-off rates
• No adaptation
• Feels robotic

Conversational AI recruiter:

• Adapts to candidate responses
• Asks follow-up questions
• Feels like talking to a person
• Gathers better information

Example:

Traditional form asks: "Years of experience?"

Our agent asks: "Tell me about your background" Then listens and adapts: "You mentioned managing a team in hospitality. How did you handle difficult customer situations?"

Better information. Better candidate experience.

The Verification Advantage

Email verification solves multiple problems:

Spam prevention:

• Bots cannot verify emails
• Disposable addresses get flagged
• test@test.com never reaches hiring managers

Quality signal:

• Verified candidates are serious
• They took the extra step
• Real email = Real person

Admin efficiency:

• See all submissions immediately (PENDING)
• Know which are verified (VERIFIED)
• AI assessment on both
• Decide whether to reach out based on verification + AI flags

Real Outcomes

For our hiring process:

• Three separate career paths (Sales, Tech, Pitch-Your-Role)
• All use same verification flow
• AI assessment on every submission
• Jesse + Caroline get job applications
• Jesse + Shawn get customer inquiries
• Zero spam reaching decision makers

Candidate experience:

• Talk to AI recruiter OR fill out form
• Verify email (one click)
• Get confirmation: "We'll be in touch within X days"
• Professional, clear, fast

Hiring manager experience:

• Immediate notification when someone applies (PENDING)
• AI tells you if they seem real/fake
• Second notification when they verify (VERIFIED)
• Focus on verified candidates only
• All info already gathered and organized

Tech Stack

Conversational AI:

• LangChain for agent orchestration
• Meta-Llama-3.1-70B-Instruct-Turbo for conversation
• Together AI for inference
• Intent recognition and routing

Verification System:

• PostgreSQL (Vercel/Neon) for token storage
• Resend API for email delivery
• Crypto-secure token generation
• Session tracking and spam detection

Application Processing:

• Next.js API routes
• FormData handling for resume uploads
• AI-powered candidate assessment
• Automated admin notifications

Frontend:

• React + TypeScript
• Cloudflare Turnstile for CAPTCHA
• Real-time form validation
• Mobile-responsive design

What Makes This Work

The combination matters.

Conversational AI alone would still let spam through. Email verification alone would feel rigid. AI assessment alone might miss context.

Together:

• Conversation gathers good information
• Verification filters spam
• AI assessment flags quality
• Humans make final decisions

For Companies At Scale

This system works for us. It would work 10x better for companies drowning in applications.

If you process hundreds of applications per week:

• Conversational screening reduces drop-off
• Email verification cuts spam by 90%+
• AI assessment surfaces top candidates
• Hiring managers see only verified, pre-screened applicants

If you have recruiting teams burning out:

• Eliminate manual spam screening
• Get AI reasoning (not just pass/fail)
• Focus time on qualified candidates only
• Track all submissions with full context

If you lose good candidates in the noise:

• Conversational experience keeps them engaged
• Fast-track option for efficient candidates
• Better information gathered upfront
• Clear next steps reduce ghosting

Implementation Options

Option 1: Full System

• Conversational AI recruiter
• Email verification pipeline
• AI assessment layer
• Admin notification system
• Integrated with your ATS

Option 2: Verification Only

• Add email verification to existing forms
• AI assessment on submissions
• Admin PENDING/VERIFIED notifications
• Keep your current application process

Option 3: Conversational Only

• AI recruiter for candidate engagement
• Routes to your existing application system
• Better information gathering
• Improved candidate experience

All options are outcome-focused. We measure spam reduction, time savings, and candidate quality improvement.

What We Learned

Conversational beats forms: Candidates engage more. They provide better information. Drop-off rates are lower.

Verification is mandatory: Without it, spam drowns signal. With it, hiring managers see only serious candidates.

AI assessment needs reasoning: "This candidate scores 7/10" is useless. "This candidate provided specific company research and clear role alignment" is actionable.

Humans make final decisions: The AI screens. Humans hire. That will not change.

Built For Us, Available For You

We built this because we needed it. Small team, multiple roles, constant hiring. Spam was wasting our time.

Now we have:

• Three career paths (Sales, Tech, Pitch-Your-Role)
• All verified automatically
• AI assessment on every submission
• Conversational option via recruiter agent
• Zero spam reaching decision makers

Your organization likely needs this more than we do.

If you process hundreds of applications per week, this system would save your recruiting team dozens of hours and surface better candidates faster.

Take the Gain, Use Your Brain

Our philosophy applies to recruiting too.

AI gets you 80% of the way (filtering spam, gathering info, initial assessment). You use your brain for the 20% (evaluating fit, making offers, building relationships).

Do not let AI make hiring decisions. Let it eliminate the noise so you can focus on what matters.

Want This For Your Organization?

We can implement this system for companies dealing with high-volume recruiting.

What we deliver:

• Conversational AI recruiter (customized for your company)
• Email verification pipeline
• AI assessment layer with reasoning
• Admin notification system
• Integration with your existing tools

Timeline: 5-10 business days depending on complexity and integration needs.

Outcome-focused contracts: We measure spam reduction and time saved, not hours billed.

Talk to us about recruiting automation

See our other case studies