Resources

What we learned putting it in production.

No conference theory: what actually breaks when an agent touches real customers — written so you can use it without hiring us.

A portal with two clients inside is no longer a website

Showing a client their data is easy. The hard part is that account A can never see a record from B — not by a bug, not by a guessed URL.

3 min read

When automation fails halfway through

Step three of five goes down, the order was created and the invoice wasn't. Nobody finds out until a customer complains. That's where a flow shows, or a system.

3 min read

AI content: writing was never the bottleneck

Publishing thirty pieces a month is easy and moves nothing. What's scarce isn't the text: it's having done something worth telling.

3 min read

Your site and your agent have to read the same catalog

If the page says one thing and the assistant another, the customer sees both on the same screen. That's where a platform shows — or two systems taped together.

3 min read

AI inventory forecasting: the data is already in your ERP

Every stockout that caught you off guard was announced in your own sales history. You're not missing data — you're missing reading it in time.

3 min read

What a WhatsApp agent actually costs (and why “$0/month” is a lie)

The numbers change by country and scope. The price structure doesn't — and knowing it is what lets you compare quotes.

2 min read

Retail demand forecasting: start with what you already have

You don't need a data lake or a six-month project. You need your sales history, twenty items, and an honest way to measure whether it's right.

3 min read

Automating operational email with AI: triage, not auto-replies

The operations inbox isn't correspondence: it's a work queue in disguise. Classifying and routing it is automatable today. Replying for you isn't — yet.

3 min read

How to test a chatbot before launch: the scenario bank

A demo that went well isn't evidence. Evidence is a bank of hard conversations the agent passes before every change — one that grows with every real mistake.

3 min read

Wiring your ERP to a sales assistant: start read-only

The system already knows the price, the stock and the size. What's missing is answers in seconds where the sale happens — writing nothing back, and never guessing when it's down.

3 min read

Supervising a chatbot in production: the panel your team needs

The agent handled sixty conversations yesterday. If knowing what it said means asking someone for screenshots, you don't have a system: you have a black box with access to your customers.

3 min read

Automating with a human gate: the AI proposes, a person signs

Useful automation doesn't remove decisions: it removes the preparation of decisions. The craft is designing what queues up for your signature — and what doesn't.

3 min read

Acquisition with attribution: which channel brings clients, and which just brings clicks

The ad agency reports clicks, content reports reach, SEO reports rankings — and nobody reports which of the three brought the client who signed.

3 min read

WhatsApp chatbots in Costa Rica: three different products with the same name

A button tree, a platform your team answers from, and an AI agent wired to your catalog all advertise the same way. The first task isn't comparing prices: it's knowing which one you're being offered.

3 min read

Cost per conversation: the unit that keeps your AI agent honest

The monthly bill says how much you spent; it doesn't say on what. The conversation is the unit that lets you compare — against last month, against the value of a sale, against a person doing the same job.

3 min read

AI ROI in retail: how to calculate it without the smoke

“Companies cut inventory 20-30%” is other companies' ROI, on other data, measured by whoever's selling. Yours gets computed from your numbers — or it doesn't get computed.

3 min read

Why AI pilots die before production

The demo impresses, the pilot gets approved, and six months later nobody uses it. It's not the model: it's the engineering missing around it.

2 min read

Anti-invention: an agent that never improvises in front of a customer

A model that doesn't know the answer fills the gap with a guess. Here's how to build the system that won't let it.

3 min read

What to check before putting a WhatsApp agent in front of customers

The short list that separates a bot that helps from one that burns customers — useful even if you build it with another provider.

2 min read