Meet Alex Sandoval, oakpool advisor and CEO and co-founder of Allie Systems, an industrial AI company. Alex spent years in senior product and monetization roles at Rappi, one of Latin America’s fastest growing consumer tech companies, before pivoting into industrial AI, building systems that run on factory floors instead of phone screens.
Alex’s perspective is a rare one. He’s built businesses on both ends of the spectrum, consumer tech and industrial operations, and brings that dual vantage point to everything he advises on, including oakpool.
1. You’ve built in hypergrowth consumer tech and now industrial AI. How has your view of AI changed from software to real-world operations?
In consumer software, AI optimizes a screen: the next tap, the next order. In operations the mistake is tangible: a stopped line, scrap, a safety event. That tangibility is what forces accountability and traceability across the board — every decision has to trace back to why it was made. And you’re not reasoning over a clean digital state. You’re reasoning over a physical, three-dimensional environment—temperature, spatial layout, geospatial conditions—that’s moving the whole time. That changes the game entirely. Consumer AI gets to be probabilistic and opaque and still win. Operational AI has to be causal, traceable, and accountable, because the mistakes are real and you can’t roll them back.
2. At Rappi, speed was everything. In manufacturing, precision matters more. How do those environments shape AI differently?
The whole move-fast-and-break-things playbook doesn’t make it in the industry. You can’t A/B a bottling line. The physics don’t move, and one bad call is three machines downstream before you’ve measured anything. Your first impression is everything, and it’s built one hundred percent on accuracy and trust. Here’s the part people miss: engineers have spent their entire careers receiving equipment built to spec. That’s the standard they already hold software to. So precision isn’t something we’re imposing on them. It’s how they’ve always worked. You win by being right, to spec, every single time.
3. Everyone talks about AI as a productivity tool. Where do you see it becoming a true competitive advantage?
The productivity framing misses it. The real shift is amplification. The people who already win — the ambition-driven, the relentless, the ones who chase a problem until it breaks — AI multiplies their impact by fifty, by a hundred. It doesn’t lift a mediocre operator to average. It takes your sharpest problem-solver and hands them the leverage of fifty of themselves. That’s where the advantage actually lives, and it compounds: the distance between a high-agency person with that leverage and everyone else widens every quarter. So the question for a company isn’t “do we have AI.” It’s “who are we amplifying — and are they our best people?”
4. You’re building AI for the factory floor. oakpool is building for the brand layer. What parallels do you see between operational intelligence and market intelligence?
The decisive reader stopped being human in both places. I’m building my company on the basis of whether a machine can understand a production line well enough to act on it. Oakpool is building theirs on whether a machine understands a brand well enough to speak for it. And the edge in both is identical: it’s built by gathering the data where you have an asymmetric advantage – the data nobody else can reach. That’s what truly makes a model the best, not the architecture. On the floor, it’s the causal history of how your specific line fails. For a brand, it’s the proprietary signal of how it’s actually understood and chosen. Whoever owns the asymmetric data owns the machine’s understanding.
5. What do enterprise buyers ask about AI today that they weren’t asking even a year ago?
The question flipped from “what can it do” to “will it behave the same way next Tuesday at 3am with a tired operator running it.” A year ago they wanted demos. Now they interrogate determinism — same inputs, same output, every time, with a trace back to the data. The ones who’ve actually run pilots learned the hard way. Enterprise buyers are buying predictability.
6. How do you think enterprise trust in AI gets built?
You build it the way you trust a person – not by being impressive, but by being predictable when it’s boring. The system shows its reasoning, behaves identically under identical conditions, and shuts up unless something is actually wrong. Every false alarm is a withdrawal from the trust account, and most AI loses its operators by week two by over-alerting. You earn the right to act autonomously by being right, visibly, when the stakes are low — long before anyone lets you touch the high ones. Trust is a deposit schedule, not a launch feature.
7. What made you want to advise oakpool.ai at this stage?
Honestly, it comes down to James [James Hamilton]. He was talking about GEO — AI-based SEO, optimizing for the engine instead of the search bar, when the category was still undefined and the only people working on it were AI Researchers. Most people are reacting to this shift now that it’s obvious. James and the team saw it coming early and built real conviction when there was nothing to point to yet. That’s the kind of founder I back. It also helps that I’m already living the same thesis on the industrial side, so I can be genuinely useful instead of decorative. The brand layer and the factory floor are running the same play. I get to pressure-test theirs and sharpen mine.
8. What industries do you think will be transformed fastest by AI over the next 3–5 years?
Follow the messy data. The fastest-moving industries are the ones drowning in unstructured information — clinical notes, contracts, claims, maintenance logs, manuals nobody ever organized — where the answer is buried in text instead of sitting in a clean database. That’s exactly what LLMs are built to read. Now weigh that against the size of the outcome. Healthcare, legal, insurance, financial services, industrial operations: the cost of being right or wrong is enormous, so every piece of buried knowledge a model can surface turns into real money made or real risk avoided. Those are the places the next wave lands first, because that’s where the value was locked up and nothing before could get to it.
