While messing around with friends in high school, Luis Loaiza built his first chatbot at 14 using IRC protocol. He's been obsessed with real-time communication ever since — connecting Blackberry with iOS and Android before that was standard, building a popular multiplayer card game called Club 40 on Facebook in the era of Zynga and Mafia Wars, even creating an encrypted email app called Cryptex in 2012 that was basically Snapchat for emails.
Fast forward to 2026 and he's building Jelou AI, a transactional infrastructure for AI applications on WhatsApp. When I sat with Luis, one thing became immediately clear: He's not interested in demos. He's interested in production-level AI that can move money, open bank accounts, and complete transactions securely, especially in emerging markets where WhatsApp is the primary digital avenue for businesses.
Here are three things I learned from Luis about what it actually takes to deploy AI that does real work.
1. Everyone Can Build a Demo, But Transactions Require Infrastructure
The demo market is huge. Today, anyone can build a demo or website; tools like Lovable, Cursor, Replit, and v0 make that incredibly easy.
But transactions? Moving money, opening a bank account, underwriting credit with a signature that's legally valid in a specific country — not DocuSign for the US, but the equivalent that's approved by the local government and regulatory entities — that's complex. The challenge is building something that works reliably at scale, in compliance with local regulations, with proper authentication and identity validation. It's taking infrastructure that was built for mobile apps and websites and making it work in the agentic world, in WhatsApp, where billions of people are already communicating daily.
Luis's company is building the operating system, what they call Brain OS, that sits on top of messaging applications and adds transactional layers. Authentication tools. Identity validation. Integrations with payment processors. Regulatory frameworks around the tools these AI applications use so companies can be in compliance.
Last year, Jelou AI executed more than $120 million in financial operations through AI applications on WhatsApp. That's proof people are using it, trusting it, and it's working at scale.
2. Enterprises Want AI, But They Want It Secure and Controlled
There's a lot of hype around AI right now. Enterprises invested in AI and startups are scaling fast. But the reality is companies need processes. They need control. They need to comply with regulations.
Luis made this point beautifully: There are billions of people with general intelligence already out there. But enterprises still need to be able to orchestrate how AI executes work and transactions. You can't just throw a task at a couple of agents, let them figure it out, and hope for the best — especially when real money and sensitive data are involved.
The challenge today is educating the market that deploying production-level AI means having everything under control. It means adding authentication layers with end users. Validating identity. Understanding if a transaction or process the AI is doing is actually approved by that user. Think of it like credit cards. We still have massive problems with fraud using conventional payment methods. We're adding a new level of complexity with agentic AI. We need guardrails, regulatory frameworks, and tools that help companies adopt AI safely.
That's the real barrier to adoption. Businesses are realizing there's enormous complexity when you want to deploy something that actually does work in production.
3. The Future of Interfaces Is Less Friction Between Humans and Computers
I asked Luis about what's coming next. His answer? Brain-computer interfaces.
Gen AI is already changing how humans connect with software. Conversational interfaces feel natural now. You can send a voice note or talk to an AI voice agent to solve something. But Luis thinks we're heading toward even more frictionless interfaces.
He referenced Neuralink and the possibility that within five years, maybe less, we'll be integrated with AI in ways where you can just think and send an email. You won't have the incentive to grab your phone, open the email app, write the email, and hit send if you can just think it. Every year, there will be less friction between humans and computers. And whether it happens with his company or someone else's, it's coming. You have to be prepared and build solutions for it.
For now, that means meeting people where they already are — like on WhatsApp, where billions of people communicate every day — and building the secure, compliant infrastructure that makes real transactions possible. Not just informational chatbots or demos. But AI that actually moves money, opens accounts, and completes work safely.
From Demos To Production
Luis's advice for anyone building in this space? Focus on infrastructure. Build for production, not demos. And remember: The real value isn't in creating one more e-commerce chatbot. It's in enabling people (not just developers, but also store owners who barely know how to set up a Shopify) to build an AI store in WhatsApp with a prompt.
That's the future he's building toward. And after this conversation, I'm convinced he's onto something that matters far beyond just another AI product. He's building the operating system for a new way of working.
Listen to the full episode of Actually Intelligent to hear more from Luis Loaiza about building a multiplayer game in the era of Facebook social gaming, why encrypted email never took off in 2012 (but the lessons still apply), and why WhatsApp is the dominant platform in Latin America for business comms.
LEAVE A COMMENT