The Right Way to Do AI Doesn’t Start With AI | Rolan Marco Garcia of Embiggen X

What we talked about.

Sunny Ray sits down with Rolan Marco Garcia of Embiggen X, a Manila-based data systems company that deploys AI for mid-market and industrial firms in construction, manufacturing and similar sectors. Garcia, a biological engineer and chemist on his third company after two bankruptcies, explains why the right way to do AI does not start with AI. He argues that broken processes cannot be automated and that fragmented data must be fixed first. He shares a hard lesson about choosing the wrong co-founder, describes a brutal facts audit that puts a dollar figure on inefficiency, and walks through how his team narrows dozens of use cases to one painful problem for a 30 day proof of value. He also explains why bespoke AI beats off-the-shelf licenses, consultants and system integrators, and why he sees physical AI, where machines connect to software for predictive maintenance, as the future of heavy industry.

You cannot automate a broken process: Rolan Marco Garcia on fixing workflows and data before any AI gets deployed in factories.

The questions, and the answers.

What is Embiggen X building and who do you serve?

We are a data systems company. We deploy AI to large companies, mid-market firms and industrial sectors like construction and manufacturing. That is our specialty, though we do work with other industries too.

What did your second failure teach you that the first did not?

Partners matter most. I was building a venture studio style fund and I picked the wrong co-founder. Integrity and values must be aligned. My partner used the funds we raised for his wedding in Morocco, and we lost money and hurt a lot of people.

Why do you say the right way to do AI does not start with AI?

You cannot automate a broken process, and AI is always garbage in, garbage out. Fix and re-engineer the process first, then consolidate your data, then put AI on top. Bolting AI onto a broken process or fragmented data means it never understands your company context.

What do clients least want to hear during the first week audit?

That what they have done for 30 years has been costing them millions. One insurance company lost at least 5 million dollars in annual revenue because manual claims processing was slow. We quantify losses from their P&L, because AI is a business value conversation, not a technology one.

How do you reach a proof of value in 30 days in heavy industry?

Clients arrive with something like 38 use cases. We pick the single most painful problem and the smallest piece we can solve, then ignore everything else. For a Swiss manufacturer, that meant automatically rescheduling production when raw material deliveries run late, connected to their ERP.

Which do you replace: consultants, systems integrators or licenses?

All of them. Consultants hand you a strategy deck, integrators stitch together off-the-shelf tools, and companies pay for whole platforms while using about 10 percent. We build bespoke AI that works how you already work, which drives adoption and ROI.

How did your first customers find you with no capital?

We are bootstrapped and customer funded, which I think is the right way, since people paying you to solve problems is the highest vote of confidence. In AI services, especially in Asia, it is all about relationships and who you know, then word spreads.

What does an industrial company look like in 10 years?

The future is physical AI. Operational technology, the machines, marries software, so you can predict failures before they happen and track every asset, like 3,000 tractors across the US, including location, fuel and maintenance needs.