Lead Generation Case Study: AI Agent Auto-Scores Every Lead

TL;DR: A real-world breakdown of how I automated lead capture, scoring, and nurturing at Toi La Tung using AI Agents — the system architecture, the logic behind it, and what I observed after running it. The system is still running in production, and deeper performance analysis will be updated each operating cycle.

The ultimate measure of an AI consulting firm's capability is not the number of certifications they hold, but whether they actually run their own business on the systems they advise clients to build (Building in Public).

In mid-2025, I ran into a real bottleneck: lead registrations on my website were scaling up, but I didn't have the bandwidth to qualify and respond to all of them in time.

When a lead registered at 10 PM, I wouldn't get to reply until 9 AM the next morning. That overnight delay cooled down genuine interest before a real conversation even started.

To fix this, I built an automated Lead Generation & Nurturing Engine powered by AI Agents. Here is the system architecture I'm currently running.

Architecture of the Automated AI Lead Gen Engine

How does the system automatically respond to leads within 3–5 minutes?

The system uses Make.com as the orchestrator: (1) It captures lead details from the landing page form, (2) An AI Agent instantly researches the lead’s website and LinkedIn profile to calculate a potential score (Lead Scoring), (3) It drafts a highly personalized email reply, and (4) It triggers an urgent notification in Slack with the complete lead profile, allowing me to approve the email or call with a single click.

Here is the data flow diagram of the system:

[Khách hàng điền form tư vấn] 
       ↓
[Make.com Webhook nhận data] 
       ↓
[AI Agent: Quét thông tin doanh nghiệp khách hàng & LinkedIn] 
       ↓
[AI Agent: Chấm điểm Lead (A, B, C) + Soạn dự thảo email cá nhân hóa]
       ↓
[Slack Alert gửi cho tôi: "Có Lead hạng A đăng ký!" + Kèm draft email]
       ↓
[Tôi bấm "Approve" trên Slack] ⟶ [Email được gửi đi tự động]

What Changed After Running This Engine

After operating this system, four things shifted in how I handle leads:

• First response speed: Leads who registered outside business hours used to wait until the next morning. Now Slack fires an alert within minutes of form submission — I just review and send.
• Pre-call research: I used to manually look up every lead’s website and LinkedIn before replying. The AI now summarizes all of that inside the draft email automatically.
• Operating cost: Tooling infrastructure (Make.com + LLM API) runs roughly $15–20 USD per month (about 400,000–500,000 VND) — far below the cost of a part-time data/admin hire (at least 5–7 million VND per month).
• Conversion: Every lead now gets a personalized reply within 3–5 minutes instead of waiting for end-of-day manual handling, so email open rates and booked consultations improved noticeably regardless of staff working hours.

Deep Dive: The Two Most Valuable Features

1. Intelligent Lead Scoring and Qualification

Not all leads are created equal. Some are students looking for templates, while others are Tech Leads or CEOs of 100-person companies.

I trained the AI Agent on my Ideal Customer Profile (ICP). When a form submission comes in, the AI automatically scrapes the prospect’s company website, analyzes their business size and industry, and categorizes them:

• Tier A: CEO/Founder of companies with >30 employees (I call within 5 minutes).
• Tier B: Department managers / SMBs (Enrolled in an automated nurturing sequence).
• Tier C: Students / researchers (Routed to my monthly newsletter list).

2. Deeply Personalized Email Drafting

Instead of sending generic template emails like "Thank you for your interest...", the AI Agent reads the prospect’s latest LinkedIn post or analyzes their current landing page to deliver a hyper-specific observation:

"Hi Nam, I took a look at your landing page for Course X. It looks like your VietQR configuration is currently hitting a CORS error when clicked on mobile devices. This is likely causing a high drop-off rate at checkout..."

Receiving an email that pinpoints a specific technical issue just minutes after signing up tends to make a genuine impression. In the cases I've observed, it shifts the conversation from cold outreach to a real technical dialogue — a much stronger starting point than a generic thank-you.

Key Takeaways from This Case Study

For SMBs, optimizing lead handling with AI doesn’t require Salesforce or HubSpot. A Make.com workflow combined with the Claude API is enough to stop missing leads that register outside business hours.

If you’re running into a similar problem — leads coming in but responses too slow, no real-time qualification — this is the architecture I’m currently running. The system is still running in production, and deeper performance analysis will be updated each operating cycle.

Read Next

• Implementing Exit Intent Popups & Lead Automation with Make.com
• The Cost of AI Adoption for Vietnamese SMBs — Calculating ROI Accurately
• Optimizing Landing Page Conversion Rates with AI-Driven Insights