ChatGPT Image Jun 11, 2025, 05_26_32 PM.png

Some problems you can ignore until they hurt. This one punches you in the face.

Reconciliation isn’t something you notice when things are going well. It’s when the numbers don’t match, when money’s missing, when you realize that nobody has a full picture of where the funds are, or where they went. That’s when it becomes real.

While building the Insurtech division at Mercado Pago, every time we launched a new payment method or brought on a new partner, reconciliation quickly became the bottleneck. It wasn’t just about confirming if we got paid. It was cancellations, refunds, voided transactions, mismatched IDs, timing differences across systems. Every day the same questions came up: Did we actually receive the money? Where is it? Who’s checking? And most of the time, no one had a clear answer. Everyone was working with partial data and hoping it added up by month-end.

In 2020, at Mercado Libre, we uncovered a major issue with Cobro Express, a cash payment provider in Argentina. They handled just 2% of our volume, but the financial impact was massive. On the surface, everything reconciled (transactions appeared valid). But when we finally reconciled what they owed us against what was actually being deposited, the gap was immense.

There was no reconciliation between the transactional and financial layers. Payments marked as completed were delayed or never settled. Based on internal reviews, some local agents were likely collecting cash and holding it back. We took a $27 million loss, recorded as bad debt.

After that, Mercado Pago built a dedicated reconciliation team, implemented Salesforce for case tracking, and changed how we worked with providers. But the damage had been done. We learned the hard way that without rigorous, real-time reconciliation, even small partners can create outsized risk.

I assumed it was a one-off. It wasn’t.

Nacho, my co-founder, was at Comun during the same period Synapse collapsed. They had to track balances across multiple banks, processors, and rails, and there was no central source of truth. Everyone was working off different data. Every week there was a new mismatch, and the team had to scramble to figure out if the problem was real or just a reporting delay. Meanwhile, Synapse (one of the core infrastructure providers in the space) imploded. Hundreds of millions went missing. Founders were left in the dark. Users couldn’t access their funds. And no one knew what belonged to whom.

It was the same problem, just at a much bigger scale.

We started Rexi because we saw the same pattern playing out again and again:

No one was building real infrastructure to solve it. Most tools are built for accountants, not operators. Others just offer dashboards that sit on top of broken pipes.

So we decided to build what we wish we had.

Rexi is an AI-native system of record for fintech reconciliation. It helps companies track, match, and understand every inflow and outflow across banks, processors, partners, and accounts. It’s built for operators who want to move fast without risking everything on the backend.

We don’t think reconciliation should be a department. It should be infrastructure. Something that works quietly in the background, and gives you confidence when things don’t go as expected.