AI agents for business

I build conversational AI agents trained on your knowledge base. They serve customers 24/7 on WhatsApp Business, qualify leads by need, budget and urgency, and only escalate to a human when there's real value.

Book a discovery call View case studies

What an AI agent does

An expert advisor that never sleeps

A well-trained AI agent isn't a generic "press 1 for sales" chatbot. It's a real expert on your product or service: it understands natural-language questions, replies in your brand's tone, knows your pricing and policies, and knows when to step back and hand off to a human.

At JL Mercantil the AI agent processes 400+ daily messages. It filters by customer type (reseller, installer, end consumer), technical need, budget range and urgency. It only hands off to the human team when there's real value — eliminating the noise of pure browsers.

The stack I use

I work with OpenAI, Anthropic Claude and local models (DeepSeek, Llama via Ollama) depending on the privacy posture the client requires. Training uses RAG (Retrieval Augmented Generation) on your real knowledge base — not generic prompts. The WhatsApp Business API and CRM integration is orchestrated via n8n.

What's included

Processes I automate

Every project is tailored to your existing channels and operational volume. These are the most requested processes:

Automatic lead captureMeta Ads, web form, WhatsApp Business and inbound calls converge into a single traceable CRM.

24/7 AI service agentBot trained on your knowledge base. Filters by need, budget and urgency before handing off to a human.

Custom CRMFull traceability from entry → classification → follow-up → close. Built on Supabase or a bespoke stack.

Automatic bookingGoogle Calendar integration so qualified prospects can schedule meetings without human intervention.

Team notificationsReal-time alerts to the responsible rep when a lead moves stages or meets a trigger criterion.

Automated reportsDaily and weekly dashboards with real pipeline metrics so decisions are made with data.

Real metrics

JL Mercantil — 400 msgs/day handled in 4 days

A technology and energy company drowning in roughly 400 daily messages from Meta Ads. I designed a bot + Supabase CRM in four days, with an n8n workflow wired to OpenRouter AI, the WhatsApp Business API and team notifications.

See all cases

JL Mercantil CRM — automation case study by Frank Cardona

Frequently asked questions

What people ask the most

What is a business AI agent?+

A conversational bot trained on your business knowledge base (products, services, pricing, policies). It serves customers 24/7 on WhatsApp or web, qualifies leads by urgency, budget and need, and hands off to a human only when there's real value.

Does it work for small companies or only large ones?+

Especially valuable for saturated SMBs. When a business handles 100-500 daily messages, hiring three more reps is expensive and inefficient. A well-trained AI agent replies in 2 seconds and only escalates the conversations worth a human's time.

Will customers notice they're talking to a bot?+

For ethics and compliance I recommend identifying the agent clearly, but the tone and answer quality keep the experience strong. The trick is that the bot resolves 70% of common questions and gracefully hands off when things get complex.

Which technologies do you use to train the agents?+

OpenAI, Anthropic Claude and, for privacy-focused clients, open-source models like DeepSeek or Llama running on local Ollama via Docker. Training uses RAG (Retrieval Augmented Generation) on your real knowledge base — no generic prompts.

Free diagnosis

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