Case JL Mercantil · 400 messages/day in 4 days
An AI bot, a custom Supabase CRM and an n8n flow on the WhatsApp Business API, live in 4 days and handling 400+ daily messages from day one.
400 daily messages, no answers
JL Mercantil (technology and energy, Medellín) was drowning in roughly 400 daily messages coming in from its Meta Ads campaigns.
The goal was for every message to be logged, for the bot to handle what it could, and for anything that needed a person to reach the responsible sales rep.
Why WhatsApp Business API and not the free app
The free WhatsApp app can't connect to a bot or a CRM. With the WhatsApp Business API, every message comes in through a webhook and is processed automatically before a person reads it.
The system that went into production
Four pieces wired into a single flow:
After the bot, a store that takes payments
The bot organized customer service; the next step was a website that sells on its own. I built the jlmercantil.com store with custom code: today it has more than 400 products and takes payments online with Wompi. The order total and shipping are calculated on the server, shipping is quoted with Envía before payment, and an order is only marked paid when Wompi's webhook confirms it. Customers can check out as guests or sign in with Google to save their address, and shipping labels are generated from the sales panel of the same CRM. A purchase is reported to Meta only once the payment is approved. The WhatsApp bot still answers questions, but the purchase happens on the website.
Here is how an online store with Wompi payments is built, piece by piece.
Welded Mesh — replicable B2B Grow Partner
Another trackable commercial system, for a construction-materials supplier in Medellín: a technical site with transactional SEO, WhatsApp routing and a Supabase pipeline, in a template built to replicate across B2B niches.
See the Welded Mesh case
About this case
JL Mercantil (technology and energy, Medellín) was drowning in roughly 400 daily messages from Meta Ads. Every message needed to be logged, and the sales team needed to receive only the conversations that required a person.
An n8n flow on the WhatsApp Business API that captures every message, processes it with AI (OpenRouter), extracts intent and lead data, logs it in a custom Supabase CRM and decides whether the bot replies or hands off to a human, notifying the responsible sales rep.
4 days. The system has handled 400+ daily messages from day one, in under 2 seconds per message.
Yes. WhatsApp Business API + AI agent + CRM + n8n is the same base I use in AI agents for business and sales automation — what changes per client is how the bot is trained and the qualification rules.
Yes. Besides the bot, I built the jlmercantil.com store with custom code: more than 400 products, online payments with Wompi, shipping quoted with Envía before payment and labels generated from the CRM sales panel. Here is how an online store with Wompi payments is built.
30 minutes. A clear map of what to automate first.
Book a 30-minute strategy call via Google Meet. You'll walk away with a concrete plan — even if we don't end up working together.
Book a discovery call