From 0 to 45 Cold-Contacted Leads in 4 Hours: The WhatsApp Lead-Gen Flow with EZContact MCP

This week we ran a real experiment: identify event venues in Monterrey, pull their WhatsApp numbers, draft and approve a Meta template, and send the first message to all of them. No SDRs. No agency. Less than a day of work.

Result: 47 leads identified, 45 with valid phone numbers, 100% delivery (45/45). Total time from “I have an idea” to “all 45 prospects have the message in their WhatsApp inbox”: 4 hours.

This article walks through the full stack, real timings, and why this pattern is replicable for any SMB vertical.

---

The problem we’re solving

If you sell B2B in LATAM, you know the scenario: you need to reach 50 prospects in the same niche (clinics, hotels, restaurants, salons, law firms), SDRs are expensive, US-built tools don’t treat WhatsApp as a primary channel, and you end up sending messages by hand from your phone.

The deeper, technical problem: you cannot message a stranger on WhatsApp cold. Meta forbids it. The only legal way to start a cold conversation is with a pre-approved Meta template (also called HSM or message template).

The real friction is: identifying prospects, drafting the template, sending it for approval, then dispatching the batch — all in a way that complies with Meta and doesn’t get your number blocked.

That’s what we automated.

---

The full stack

1. Lead scraper in Python with Playwright

A Python script using Playwright that takes a query (e.g., “event venues in Monterrey”) and crawls Google Maps + Google Search. For every business, it extracts:

• Name
• Phone (including WhatsApp flag)
• Linked Instagram
• Rating and reviews
• Address

Output: a clean TSV ready to use.

Real time for 40 businesses: ~4 minutes.

The key thing about this step is that the scraper is vertical-agnostic. Switch the query to “dental clinics in Guadalajara” or “corporate law firms in Mexico City” and you get the same output format. One tool, infinite niches.

2. Google Sheets as a dashboard

The TSV uploads to a Google Sheet with tracking columns: contacted, replied, interested, booked. Nothing fancy — just a simple panel where the operator sees what leads they have and what’s happening with each one.

3. EZContact MCP (JSON-RPC)

This is where EZContact comes in as a programmatic tool. The EZContact MCP (Model Context Protocol) exposes JSON-RPC functions for:

• Creating a WhatsApp template with the copy and buttons you define
• Submitting it to Meta for approval
• Sending the batch to a phone list with configurable delay
• Tracking of delivered, read, and replies

In this run we used a 3-second delay between messages to be polite to Meta (Portfolio Pacing is active in 2026 — best not to push limits).

Meta template approval time: ~3 minutes. Batch send time (45 messages with 3s delay): ~2.5 minutes.

4. The EZContact AI agent takes over

This is the part that changes the game. When a prospect taps any of the template buttons (“Tell me more,” “What’s the price?”, “Not now”), the WhatsApp 24-hour service window opens and the EZContact AI agent enters the conversation.

The agent was configured with a single prompt describing the business, prices, common objections, and the goal (book a demo). No 40-node n8n flows. No decision trees. Just a plain-language prompt.

The agent:

• Replies with prices instantly
• Captures prospect data
• Filters serious leads from tire-kickers
• Books the demo on the calendar
• Activates the account if the prospect wants to start

The human at the business does not intervene in the first conversation. They only receive leads that are already filtered and ready to close.

---

The template we sent

This is the exact message Meta approved and that all 45 venues received:

Hi 👋 I have something I think might work well for [Venue Name]: an AI agent on WhatsApp that responds instantly to your customers, filters serious leads, and books visits 24/7 — so you don’t lose a single prospect even when you’re in a meeting or off-hours. Setup takes less than 5 minutes. Want me to tell you more? — EZContact.ai

Buttons: Tell me more | What’s the price? | Not now.

Three things make it work:

1. Variable personalization — the venue name is injected per contact. Meta allows this with {{1}} and it boosts engagement.
2. Concrete benefit, not generic — “you don’t lose a prospect when you’re in a meeting” is the real pain of a venue owner.
3. Low-friction buttons — the prospect responds with a tap, not a typed reply. The “Not now” matters: it cleanly removes uninterested contacts without burning the relationship.

---

Real timings

Step	Time
Idea + brief	30 min
Scraping 40 businesses (47 leads)	4 min
Phone validation and cleanup	20 min
Template draft + Meta-friendly tweaks	45 min
Submit + Meta approval	3 min
Batch send (45 messages, 3s delay)	2.5 min
AI agent setup for handoff	30 min
Post-send verification	15 min
Total idea → first message in inbox	~4 hours

Quick comparison: a full-time SDR takes a week to do this same work manually — researching leads, finding numbers, writing personalized messages, sending them one by one, and handling early replies. With this flow it’s 4 hours, and the AI handles replies.

---

Why it works

The scraper replicates to any niche

Change the query and you get leads for any SMB vertical:

• Dental clinics in Guadalajara
• Boutique hotels in Tulum
• Corporate restaurants in Mexico City
• Spas and beauty centers in Mérida
• Law firms in Monterrey

Same technical template, different input.

The Meta template is the only legal way to do WhatsApp cold outreach

This isn’t optional or a hack: Meta requires an approved template for the first contact with any number that hasn’t messaged you. Try to bypass it and Meta degrades your Quality Rating, eventually blocking your number. EZContact handles approval, versioning, and tracking without you touching JSON.

Buttons remove friction

A prospect tapping a button in 1 second is 5x more likely to convert than one who has to type a reply. WhatsApp templates with buttons see response rates that are 2x higher than plain-text templates in our tests.

Transparent AI→human handoff

Once the prospect replies, the AI agent runs the conversation for the 24-hour window. The customer doesn’t notice they’re talking to an AI — the conversation flows naturally. If something needs a human (negotiation, complaint, edge case), the agent proactively flags the team, who can take over without the customer noticing.

This is what differentiates an AI agent from a chatbot: the chatbot transfers and disappears. The agent stays, supporting the human in real time.

The business owner doesn’t touch the first conversation

That’s what makes the flow scale. The venue owner only sees filtered leads — those who already replied, asked about price, gave their data, and are ready to book. The tire-kickers are handled by the AI and never reach the human’s inbox.

---

What’s next: response data

Sending 45 messages with 100% delivery is half the story. The other half is what prospects do. Within 24-72 hours we’ll have:

• Open rate (Meta reports it within hours)
• Button-tap response rate
• AI-driven conversations
• Leads that reach “book a demo”

Real numbers will be published as a follow-up. If it worked, the pattern is replicable and cheap. If it didn’t, the learning is also valuable.

---

Who this flow is for

If your business hits these, the flow works for you today:

• You receive (or want to receive) leads via WhatsApp
• Your product/service sells to other businesses or to consumers searching on Google
• Your team gets overwhelmed when many messages arrive at once
• You lose leads because you can’t reply fast enough (especially off-hours)

You build the stack yourself, or we build it with you. The EZContact MCP + AI agent part is ready to use — you just need an active WhatsApp Business API account.

---

How to start

1. Pick your niche. A specific vertical works better than “everyone.” “Event venues in MTY” beats “businesses in general.”
2. Scrape or source the leads. If you don’t want to build the scraper, we can share the Python base we used.
3. Activate EZContact (app.ezcontact.ai/signup.html) and configure your AI agent with one prompt.
4. Create your first template in the visual editor or via MCP. Send for approval.
5. Run the batch with reasonable delay (3s works). Let the AI handle replies.

Want to see the full flow in a 20-minute call? Message us on WhatsApp. We’ll walk you through the first batch live.

---

FAQ

Is it legal to message strangers on WhatsApp? Only with a Meta-approved template and respecting spam policies. EZContact manages approval; you write the copy.

How many messages can I send per day? In 2026, once your business is verified, you skip the intermediate tiers and jump straight to 100,000 messages per day. More in our WhatsApp Business API Updates April 2026 guide.

Do I need a developer to use the EZContact MCP? No. The MCP is for automated flows, but anything it does can also be triggered from EZContact’s visual panel. The MCP only matters if you’ll run this repeatedly.

What if the prospect replies after the 24-hour window? Your only way to re-contact them is another approved template (cost: a new conversation). That’s why AI handoff is critical — it maximizes the window while it’s open.

How much does it cost? A marketing conversation in Mexico in 2026 costs roughly $0.04 USD. Sending 45 messages ≈ $1.80 USD in Meta cost. EZContact does not add markup on top of Meta.

---

Related resources

• WhatsApp Business API: Complete 2026 Guide
• WhatsApp Business Templates: Complete Guide
• WhatsApp API Updates April 2026
• WhatsApp BSP Comparison 2026

---

Want to replicate this flow in your niche? Talk to us — we help you build it from scraper to AI agent in a week.