AI Product · Fractional CPOUnder NDA

Turning a feature factory into a product that compounds.

A growing startup had momentum but no coherent design language or research practice. I built both, and kept shipping velocity high with AI in the loop.

Year
2024 – Present
Role
Fractional CPO
Product & design leadership
Client
AI productivity · LATAM
Founders + engineering
The Problem

Features shipped, but the experience didn’t compound.

The product had real traction signals, but releases went out without a shared design language or a systematic way to learn from users. Each release felt like a separate thing, and users weren’t building trust in the product as a whole.

The Solution

Make every release add up.

Two foundations laid in parallel, without slowing shipping: a lightweight design system for consistency, and a repeatable practice for gathering and acting on customer signal.

01
A lightweight design system
Enough structure to make the product feel like one product, not ten features.
02
A research and feedback loop
A repeatable, low-overhead way to capture customer signal and route it into the roadmap.
03
AI in the loop
Kept velocity high while raising the quality floor.
04
Roadmap from evidence
Shifted decisions from loudest-opinion to clearest-signal.
Background

This was a different problem from the fintech. Not survival, but coherence. The company had momentum and was shipping fast, but speed without a system had started to fragment the experience.

The job was to install design maturity, a shared language and a feedback habit, without putting a brake on the velocity that was working. Knowing how much process to add, and when, is core founding-designer judgment.

Research

Standing up a research practice from near-zero.

I stood up a research practice from near-zero: a lightweight, repeatable way to capture customer signal and feed it back into product decisions. The aim wasn’t a heavyweight research org. It was a habit the team could sustain, so the product kept learning without a dedicated researcher in the room.

Primary user: the productivity tool’s core user, kept generic for the NDA. The design priority: make a fast-growing product feel coherent and trustworthy as it added surface area.

How Might We…

…make every release add up to a product users trust, without slowing the team down?

Ideation

Consistency is a feature.

A coherent design language and a steady feedback loop turn scattered shipping into a product that earns trust over time. AI tooling meant the team didn’t have to trade speed for that coherence. They could have both.

Design

I built a lightweight design system, tokens, components, and patterns, sized to the team rather than over-engineered. The move that mattered was rolling it out without halting feature work. The system had to absorb the existing product and guide new work without becoming a bottleneck. That’s a leadership decision as much as a design one, and it’s the kind of call a founding designer owns alone.

Testing & Iterations

The feedback loop started changing decisions almost immediately. Signal that used to get lost now had a path into the roadmap, and the team began iterating on evidence rather than opinion. The design system made each new feature cheaper to build consistently, so consistency stopped being a tax and became the default.

Results

The product became more coherent, and the team shifted to evidence-led decisions. New features now slot into a shared system instead of arriving as one-offs, and the experience started to compound, each release reinforcing the last.

Final Thoughts
What I learned
+ Genuine lesson about installing design maturity in a fast startup.
Next steps
+ Where it’s heading.
Most proud of
Giving a team a way to iterate on evidence instead of opinion, and watching the product start to feel like one coherent thing.
Next project →
Scalable Path
Like what you see?

Let’s build something together.

felipe@felipelebrun.com
© 2026 Felipe RestrepoLinkedIn ↗Behance ↗Remote · Colombia / France · EN · FR · ES