Back to Blog
Case Studies

Stopping Fake Leads with Agentic Validation on Our Website

How we added an explainable agentic layer to our Contact page to filter gibberish and low-quality submissions without blocking real prospects.

August 28, 20252 min readBy Jesse Alton
Originally published on Virgent AI Case Studies

Stopping Fake Leads with Agentic Validation on Our Website

Industry: AI Consulting

The Spam Problem

Like many B2B companies, we were drowning in fake contact form submissions. Bots, spam, and low-quality leads were clogging our pipeline and wasting sales time.

Traditional solutions (CAPTCHAs, honeypots) weren't enough. We needed something smarter.

Our Agentic Solution

We built an AI agent that evaluates every contact form submission in real-time:

What It Checks

  1. Content coherence - Does the message make sense?
  2. Business context - Does this look like a real business inquiry?
  3. Behavioral signals - How was the form filled out?
  4. Cross-reference data - Does the email/company exist?

Explainable Decisions

Every decision comes with reasoning:

PASS: "Legitimate inquiry from verified company domain. 
      Message shows clear business need. 
      Form completion pattern matches human behavior."

BLOCK: "Message contains gibberish text. 
       Email domain is known spam source. 
       Form completed in 0.3 seconds (bot-like)."

Implementation Details

The Flow

  1. User submits form
  2. Agent evaluates (50-200ms)
  3. Pass → Goes to CRM
  4. Uncertain → Goes to review queue
  5. Block → Soft rejection with alternative contact

Technical Stack

  • LangChain for agent logic
  • Vercel Edge Functions for speed
  • Custom scoring model
  • Logging for continuous improvement

Results

Before Agentic Validation

  • 70% of submissions were spam/low-quality
  • Sales spent 2+ hours/day qualifying leads
  • Legitimate prospects sometimes lost in noise

After Implementation

  • 95% spam blocked at form submission
  • Zero false positives on real business inquiries
  • Sales time reclaimed for actual selling
  • Better prospect experience (faster response)

Key Learnings

  1. Explainability matters - We can review why any decision was made
  2. Soft rejections work - Don't just block, offer alternatives
  3. Continuous improvement - Agent gets smarter with feedback
  4. Speed is critical - 200ms budget for great UX

Try It Yourself

You can see this system in action on our Contact page. Submit a real inquiry and experience the difference between an AI-qualified lead process and traditional forms.


This is a "dogfooding" case study—we built this for ourselves and now offer similar solutions to clients.


Originally published on Virgent AI Case Studies

📍 Originally published on Virgent AI Case Studies
Share:
JA

Jesse Alton

Founder of Virgent AI and AltonTech. Building the future of AI implementation, one project at a time.

@mrmetaverse

Related Posts

Case Studies

How We Built Our Own Website with Agentic Design

A meta case study: How Virgent AI uses agents, embedded prompts, and agentic design principles on our own website to optimize inbound lead qualification and customer experience.

Case Studies

Peake.ai: We Built Our Own AI Phone System in 1 Hour

Sick of overpaying for clunky VoIP services, we coded our own AI-enhanced phone system. V1 was live in 60 minutes. LangChain automations followed an hour later. Today it powers our outbound calling.

Subscribe to The Interop

Weekly insights on AI strategy and implementation.

No spam. Unsubscribe anytime.