From Payments Data Chaos to Agentic Commerce: A Founder's Vision with Roberto Kafati, Co-Founder & CEO at DEUNA

What if your checkout button did more than collect money?

Imagine a single layer of software that not only finishes transactions but reads the customer, predicts a problem, and fixes it before revenue slips away. That’s the provocative claim Roberto Caffati makes about commerce today — and it stuck with me long after the conversation ended.

From a Honduran kitchen table to global payments

Roberto's origin story feels refreshingly grounded. Growing up in Honduras, where remittances make up a huge portion of the economy, he watched payments act as lifelines — and as bottlenecks. His early interest in crypto collided with a pragmatic realization: compliance, regulatory know-how, and institutional trust mattered. So instead of rushing to idealize blockchain, he learned the plumbing of payments at McKinsey and then built a product that merchants actually needed.

There’s something quietly heroic about that shift — from technologist to pragmatic operator. It’s the difference between an idea and an enduring company.

Why the checkout is suddenly strategic

Most people think of checkout as a final step. Roberto pushed me to see it as the richest data source a merchant has. Because every transaction aggregates identity, behavior, marketing signals, and payment outcomes, a modern checkout is less of a button and more of a sensor network. Dauna’s engineers now extract hundreds of data points per sale — and this is the leverage point for smarter commerce.

Honestly, I didn’t expect the number to land so high: over 600 data points per transaction. That scale turns the checkout into an oracle.

Agentic commerce — a new category

Here’s what stood out about agentic commerce: it’s not reactive analysis. It’s proactive orchestration. Instead of surfacing a report two weeks after a failed payment, an agentic system can detect a declining acceptance rate, surface the opportunity, and suggest or execute a fix in real time.

That matters badly for enterprises operating across dozens of countries. Human teams can’t see every fraud spike, currency glitch, or provider cap across every market. Software that understands business context, KPIs, and the right corrective moves will win.

How data becomes AI-ready

Big language models and enterprise AI projects often falter on the same problem: data preparation. Roberto’s argument is simple and convincing — capture clean, contextualized signals at the source. When data arrives already labeled with identity, intent, and payment signal quality, you can build multi-agent systems that reason about customers and revenue without months of costly engineering.

It’s a small structural pivot with huge implications: the easier the data intake, the faster merchants can deploy intelligence that actually moves the needle.

Resilience, small wins, and a two-year horizon

Beyond tech, Roberto kept returning to human themes. Resilience is his north star. Not the dramatic, mythical kind — but the slow, steady accumulation of small, compounding wins. Walks that become workouts. Iterations that become product-market fit. Minor improvements that, stacked over months, remake culture and outcomes.

I found that candid and useful. It’s easy to fetishize moonshots. Less sexy — but often more effective — is a two-year plan and the discipline to sacrifice short-term comforts for long-term position.

The most vivid moments

• Checkout as a sensor: Thinking of payments as the central data pipeline reframed everything.
• 638 data points: The scale of per-transaction telemetry is both surprising and convincing.
• Agentic systems: Automation that detects and acts on revenue leakage in real time is a genuine leap.

What founders can steal from this playbook

Start by asking pragmatic questions: what part of your product is already touching customers more than any other? For many businesses, it’s payments. Then ask: can that touchpoint feed intelligence back into the system? If the answer is yes, prioritize product work that makes the data accurate, complete, and immediate.

And on days when the work feels impossible, remember Roberto’s rule: small consecutive wins. Focus on what you can improve today. Compound that for two years and the landscape changes.

A closing thought on invention and humility

There’s a tempering humility to what Roberto builds. He’s not promising magic; he’s promising a better way to see and act on commerce. That blend of ambition and pragmatism is the kind of work that redefines categories — quietly, insistently, and with a long-term plan.

It makes you wonder: what other everyday touchpoints are hiding strategic data we’ve been treating as mere infrastructure?