Build the rails before the fast train arrives

If you're waiting for AI to "mature" before you start, you're already late. Not because you need to plug in AI right now. Because rails are laid before the train passes through, not during.

I'm speaking to SMB owners. To those of you who've spent months watching the same chorus repeat everywhere: "implement AI now", "jump in or get left behind", "automate or die". That narrative sells fear. And like every narrative that sells fear, it's incomplete.

The truth is both calmer and harder. AI will keep changing. Models will keep improving every quarter. The ones winning today aren't those who plugged in the trendiest chatbot, but those who built an operation that can receive any AI — today's, and the one coming next year.

Think of it as a fast train and some rails.

The train is AI. It moves fast, changes models, changes providers, changes pace. OpenAI this month, Anthropic the next, an open-source model after that. What looks definitive today is old in six months.

The rails are your infrastructure. Your systems, your data, your processes, your KPIs. That doesn't change every quarter. That's your business operating.

Without rails, no matter how fast the train comes, it doesn't pass through your company. You've got nowhere to put it.

What are the rails, specifically?

When an SMB owner asks me "OK Víctor, what are the rails, specifically?", I answer with five things.

First, how digitized your company actually is. If your operation still lives on scattered WhatsApps, shared Excels, and people's memory, you don't have rails. You have chaos wearing a process costume.

Second, stable systems that do the basics well. A CRM someone actually keeps up to date. Client data in one place. A documented flow of how a client enters, gets attended to, and gets invoiced.

Third, orderly data. I'm not talking about big data or complex architectures. I'm talking about whether, if I ask you how many new clients you got last month, you can answer in thirty seconds — not ask your team and wait a day.

Fourth, measured KPIs. This is the one that most separates those who are ready from those who aren't. A KPI is a number. But to have that number with confidence, you already did all the previous work: you defined what to measure, connected the sources, ordered the data. If you don't measure KPIs today, you don't have rails. No matter what anyone tells you.

Fifth, centralization and real time. That you, as owner, can see the state of your operation without having to gather three people to tell it to you.

That's the rail. Easy to say, expensive to build.

The Travel Remesas case

I see this pattern in almost every first conversation with SMBs. They arrive asking about AI, but when you scratch the operation, they don't even have the basic rails.

Travel Remesas came to me with this situation a year ago. Their operations lived on WhatsApp. They had started using Odoo, but Odoo wasn't adapted to their business. What they had was many layers of abstraction on top of a generic tool. Outdated database. Someone writes on WhatsApp, then someone else passes the data into Odoo by hand, then Odoo fails to report. No up-to-date KPIs. No real-time data.

They wanted to plug in AI. They'd already heard it was the moment.

I told them no. Rails first.

We tore down the layers. We designed a platform built for their specific operation. We put user data into a structured database. We defined the metrics that actually mattered for a financial operation. We built the dashboard.

Today Travel Remesas processes over 400 daily operations with over 1,500 active users and full visibility of their business. They can now start plugging AI into specific parts: customer support, filtering, suspicious-operation detection. Because they have somewhere to plug it.

A year earlier, they would have spent thousands on an AI agent that would have had nowhere to connect.

AI is the brain. The rails are yours.

What the rush-sellers don't tell you:

When your own systems are working, changing AI models is literally swapping a URL or flipping a parameter. This month you use one model, next month another, the month after that an open-source one you run on your own server. The brain gets swapped. The rails stay.

Whoever marries a specific AI today because "it's the best right now" will be tied to that decision in two years. Whoever builds the rails first will be able to change the brain whenever it suits them, painlessly.

That's operational freedom. That's not depending on Google, OpenAI, or Microsoft. Your business stays yours; the model is just a tool.

Why start now and not in two years?

The honest question becomes: why start now and not in two years when "everything is more mature"?

Because whoever starts now has a two- or three-year head start. Not a head start on AI. A head start on operation. When AI is ready to be seriously applied in SMBs, it won't matter who has the latest model. It'll matter who has the orderly operation to actually use it.

And building rails takes time. It's not done in a sprint. It's digitizing processes that have been manual for years. It's retraining the team to use a new system. It's migrating data. It's defining KPIs, testing them, adjusting them.

That doesn't get solved in a quarter. That's why whoever doesn't start now won't have rails in two years.

The expensive part isn't AI

You'll notice something in everything I've written: I'm not telling you to implement AI today. That isn't the stance.

The stance is: prepare a serious digital strategy. One of the services I offer at Wisd is literally called Digital Transformation Plan exactly for this reason. It doesn't start at "let's plug a chatbot in for you". It starts at "let's understand what you have, what you're missing, what rails we need to build before thinking about anything else".

Implementing AI without rails is burning money. Building rails first is making sure the money you'll spend on AI later actually pays off.

The expensive part isn't AI. The expensive part is having nowhere to put it.

---

If you're reading this and you run an SMB, take a moment to think about your real operation. Not what you tell investors. Not what it looks like from the outside. What actually happens on a random Tuesday at eleven in the morning at your company.

How many processes live on WhatsApp? How much data lives in people's heads? How many KPIs could you answer in thirty seconds?

That's what tells you where you stand with the rails.

And the train, I'll tell you calmly, is already leaving the station.