Rebuilding a design org around AI

New-feature delivery went from 2-4 weeks to 1-2, with the same human design and code gates. Screens from days to hours. 9 of 9 designers adopted the model in 2025, including its strongest initial skeptics.

Role: Design Manager, senior design leadership of the company
Team: 9 product designers across 3 squads in 2025; 3 designers today
Period: July 2025 to present
Context: meutudo, a Brazilian credit fintech serving millions of customers: INSS and private payroll loans, FGTS anniversary withdrawal, and insurance
What I personally did: the diagnosis and the operating model; the pitch that convinced leadership; the platform that makes the model possible, which has its own case on this site; and the change management, one conversation at a time.

The thesis

Code is just another prototyping tool, the way Figma is and paper once was. And drawing screens was never the job. The designer designs the experience and owns the patterns, defining them, evolving them, and knowing when and how to break them. AI is excellent at creating what already exists. So I taught it our standard, and freed the team for the part that does not exist yet.

The situation

Nine designers stretched across too many squads, with operational work eating the hours real design needed. A new feature took 2 to 4 weeks from concept to deploy. The queue kept growing, and the obvious answer, hiring more people, would not change how the work happened. In a regulated market, speed without quality is a liability, not a win.

Two models compared. In the old one, product or business decides the solution and design replicates the screen. In the new one, a request arrives and the first question is how defined it already is. If the problem is still open, design shapes the solution. If a solution arrived pre-shaped, design refines it and gives it coherence. Both paths end at the same place: validated with evidence. — DIAGRAM · WHERE WORK ENTERS, BEFORE AND AFTER

Convincing

The diagnosis I took to leadership fit in one sentence. What slows the team is not talent or effort, it is the operating model. I pitched the answer as an eleven-slide plan, with minimum governance, success metrics defined upfront and ninety days to roll out. The initiative was mine, from diagnosis to plan, and it was approved.

Journey: Each designer owns a journeyrather than handling tickets
Mission: One mission with a clear indicatoralongside a controlled backlog
AI: Does the first passalways supervised by whoever owns the journey
Bar: Success metrics defined upfrontminimum governance, ninety days to roll out

Building

For AI to execute inside our standard, the standard had to become something a machine reads. I built that part myself, with contracts, guardrails and automated review. But the tools are only the means. This case is about the mindset they made possible, where anyone can materialize an idea on-pattern, drift is caught by tooling instead of by memory, and the design and code gates stayed human.

The platform has its own case: the design system machines can read →

A phone screen drawn by hand on lined paper: the title Consignado CLT, the question how much do you need with an empty field, a row of installment options such as 55x, 60x and 48x, a checkbox for a protected loan with and without insurance, and at the bottom the interest rate, the installment amount and a Continue button. — FIGURE · FROM PAPER TO THE DESIGN SYSTEM, PRIVATE PAYROLL LOAN SIMULATION

The AI's working notes: a two-column table mapping each element of the sketch to the design system component that renders it, such as the amount input with supporting text, the installment carousel, the protected-loan toggle and the interest rate and installment summary, noting which ones still live in the legacy library. — FIGURE · FROM PAPER TO THE DESIGN SYSTEM, PRIVATE PAYROLL LOAN SIMULATION

It opened the same door from both sides. People who had never drawn a screen, and would not touch Figma out of fear, got a safe path to materialize ideas on-pattern. Designers, who master Figma but would not leave it on their own, got bridges into code starting from their own comfort zone.

Eight steps read top to bottom. Before the line: the problem, then ideation as intent and a sketch from anyone. On the line: AI raises the work to full fidelity, so the prototype is already UI code, at medium trust. Then a human design gate, where a designer decides whether problem and solution make sense. Then an automated gate where AI itself checks adherence to the design system, UI and tone of voice. Then AI writes logic and integration, at high trust. Then a second human gate, where AI and an engineer validate states and edge cases. Only then, production. — DIAGRAM · THE LINE AND ITS THREE GATES

Adoption

The hardest resistance came from inside. Three fears, in their words, and what I answered.

Am I supposed to become a programmer?
No. Technology has always been part of digital design. Code is a prototyping tool, not a career change.
You are looking at technology, not at design.
I am looking at the designer's time. If AI creates well what already exists, that time moves up: from drawing screens to deciding what should exist.
We are handing our design to other people.
We hand over the execution, not the design. Aesthetics run inside a system we designed, govern, and are the only ones who know when to break.

Arguments alone convince nobody. The team changed its mind when the first deliveries took days instead of weeks, using tools it already knew. Today the designer who had said "these things do not mix" is among the people who deliver the most.

Two timelines comparing the delivery line before and today: the earlier flow runs from problem through manual design, refinement, development, validation and production in two to four weeks; the current flow adds a design gate and a code gate, both decided by humans, and lands in one to two weeks. — DIAGRAM · DELIVERY LINE, BEFORE AND AFTER

Results
Results	Before (Jul 2025)	Today
New feature, concept to deploy	2-4 weeks	1-2 weeks max
New screen	3-5 days	under 1 day
Prototype	disposable artifact	becomes production code
Handoff	images, PDFs and Figma	usable code
Design team	9 designers	3 designers
AI adoption on the team	none	9 of 9 designers in 2025, all committing to production

Limits and learning

In 2026 the company reduced the design team from nine to three as part of its AI strategy. That was the company's decision; shrinking the team was never the model's goal. My part was to redistribute the journeys and take the team out of screen execution, so three people could design experiences and patterns for the whole company. Delivery stayed faster than in 2025, with the same human reviews.

Part of what this pipeline solved in 2025 now ships inside Figma. What did not come out of the box was the operating model and the gates, and that is what this case documents.

The model has deliberate borders. User research stayed human, AI is always supervised, and critical decisions belong to people. The lesson I carry forward is that fear only gave way after visible proof, when the first delivery landed in days. If I did it again, I would plan that first delivery for week one, before any slide.