Roger Sborchia

AI hacking with a winning hypothesis

One week, a team of five and AI on every front. The bet: a single hypothesis.

Desenvolve AI · Hackathon · 2026

In one week, as a team of five, we designed, built and presented the new Portal do Carnê: a simplified journey that offers credit and shows what is being paid, in a few clicks. With AI on every front: prototype and handoff in Figma, insights, native iOS app, back end and storytelling.

1st place overall and in ideation and prototyping (Figma)

What I did

  • Ideation and business hypothesis
  • Figma prototype with AI
  • Handoff that became code (Figma MCP)
  • Pitch storytelling
Event
Desenvolve AI Hackathon, with a jury and the Figma team
Project
Portal do Carnê
Format
1 week, team of 5
Role
Product Designer: ideation, prototype, handoff and pitch
  • 1 week from problem to working portal
  • 5 people: front end, back end, tech lead, PM and design
  • 6 fronts with AI, from prototype to pitch

One week, five people and a rule that changed everything: the most important criterion was not the prettiest or the most technical. It was delivering results for the company.

The pain we chose

In the apps, anyone looking for the Carnê landed on a webview, a website inside the app, broken and not user-focused. The path was hidden in the Account menu, and people who just wanted to understand the Carnê ended up in a payment flow. Customers did not know how to sign up or how to pay their installments.

Before: the Account menu and the webview inside the app
Before: the Account menu and the webview inside the app

And it was not a few people: the portal had a high monthly volume of visits (the exact number is confidential), and 33% of them were looking for credit.

A single hypothesis

A week does not fit everything, so we bet on one thing: take the Carnê out of the webview and turn the portal into an in-app journey that offers credit and shows what is being paid in a few clicks. Anything that did not serve that bet was left out.

AI on every front, with people deciding

I prototyped in Figma with AI while the team was building, without waiting for the layout to be done. The handoff was a method of mine that the Figma MCP read very well, so the code came out faster and closer to the design. The native iOS app, the back end, the insights and the pitch script were also built with AI, always with someone reviewing.

What we built

The new Portal do Carnê and a more inviting Account menu, highlighting the company's own payment methods. Along the way, we fixed issues nobody knew about: in the webview, customers had to log in again and, in some cases, could only get back home by closing the app.

After: the new Account menu and the Portal do Carnê journey, end to end
After: the new Account menu and the Portal do Carnê journey, end to end

The result

The expected impact was business: fewer defaults, customers paying overdue installments and discovering how to buy and check their own credit. That is what the jury rewarded.

Recognition

  • 1st place overall: Best solution award: balancing business and users, solving a real pain
  • 1st place in ideation and prototyping: Awarded by the team representing Figma at the event
The final pitch at the event, with the jury and the Figma team
The final pitch at the event, with the jury and the Figma team

AI is a multiplier, not a shortcut: it brought speed because judgment came with it.

Tags: Hackathon, Generative AI, Figma, Prototyping, Handoff, iOS, Credit, Storytelling

Keep exploring