PRV End-to-end product · sole human · built with AI

Precivox

From the product study to Precivox: design + AI so the shopper truly pays less — and the market competes with intelligence.

  • Product AI product · B2C + B2B
  • Role Product Designer & AI Builder
  • Timeline Precinho study → live product
  • Stack Figma · LLMs · Product building
  • Context PRECIVOX Tecnologia Ltda.

End-to-end product live — the only human in the process, built with AI

Open product

PRVThe system

One product, two sides.

  1. Shopper
    • builds the list
    • compares neighborhood markets
    • price, distance, time, and promotions

    Decision: net savings

  2. Market
    • registers the catalog
    • shows up in the comparisons
    • local demand becomes visible

    Decision: competes with intelligence

Precivox no celular

The shopper builds a list and compares neighborhood markets; a market can show up in the comparisons. AI at the core of the proposition, not as decoration.

Context

Precivox came out of the Precinho study: can user-centered design + AI make the supermarket shop stronger? Today it is a live platform for intelligent price comparison — the shopper builds a list and compares neighborhood markets; a market can show up in the comparisons.

Problem

The shelf price is rarely the real saving: distance, time, and promotions change the math. On the market side, there is no view of local demand that does not "burn margin".

My role

I was the only human in the process. I built the product end to end — discovery, design, product, and implementation — with AI as an execution partner.

Decisions

We saw that comparing price alone creates a false saving → net savings (the neighborhood context). We saw two audiences → a two-sided product. We saw that a study alone does not prove AI Builder → ship Precivox.

Solution & artifacts

Screens and project details.

A two-sided product, live

The solution is Precivox itself: intelligent price comparison in the neighborhood for the shopper, and a surface where a market can show up in the buying decision — with AI at the core of the proposition, not as decoration.

  • The shopper's list → a comparison with real context (not only the shelf price).
  • Local markets compete with intelligence, without "burning margin" blind.
  • Shipping the Precinho study → a live product, as proof of AI Builder.
Precivox on a phone, with the headline about paying less at a market near you
precivox.com.br on a phone — a capture of the live product.

Open precivox.com.br

From Precinho to Precivox

Precinho was the study: a design + AI hypothesis for supermarket shopping. It was not the final case — it was the lab. Precivox is what happened when the decision stopped being “polish the study” and became “ship the product”.

  • Concept that stayed — shelf price ≠ real savings; the math needs distance, time, and promotions.
  • Concept that grew — from one side (shopper) to a two-sided product (shopper + market).
  • What did not come along — study-test metrics (e.g. time percentages) without revalidation on the live product.
  1. Precinho (study)
    • UX + AI hypothesis
    • shelf-price savings
    • portfolio prototype

    Decision: prove the problem, not the product

  2. The turn
    • net savings
    • two audiences
    • explainable AI

    Decision: a concept that can carry a product

  3. Precivox (live)
    • shopper + market
    • AI at the core
    • PRECIVOX Tecnologia Ltda.

    Decision: an end-to-end product, live

What we discarded

Problem

The shelf price is rarely the real saving: distance, time, and promotions change the math. On the market side, there is no view of local demand that does not "burn margin".

Discarded

Treating Precinho as a final case of an "endless study". The decision was to ship: turn the hypothesis into a two-sided product (shopper + market) instead of polishing one more portfolio prototype.

Chosen

We saw that comparing price alone creates a false saving → net savings (the neighborhood context). We saw two audiences → a two-sided product. We saw that a study alone does not prove AI Builder → ship Precivox.

Result

End-to-end product live — the only human in the process, built with AI

Impact

Impact that can be published honestly: the product itself, live (precivox.com.br), with AI at the core of the proposition — not invented business metrics.

What I learned

AI Builder is not generating a mockup from a prompt. It is being accountable for the end-to-end outcome and using AI to think, build, and put intelligence in the product.

RRContact

Let's talk.

Tell me the context — the message comes straight to me.

What should we talk about?

Context, the problem, and a timeline, if you have one.

Prefer email? contato@romulo.art.br