Cases · live proof

Watch the AI work — and the craft that holds it up.

Two tracks, kept separate on purpose: AI that does real work, and product with engineering craft. No mislabelling, no inflated metric.

AI track — intelligence that executes

AI with software discipline: it proposes, the rule disposes. Validation, deterministic fallback and auditing behind every decision.

AI · ClaudeNévoaIA's own system.

The problem

A lead comes in and support doesn't scale — manual, slow qualification, and an opportunity that cools off.

How it works

Reads the message and scores the lead across multiple dimensions.
The AI proposes the action; a deterministic rule layer decides — with a human handoff.
Deterministic fallback when the model fails, a retry queue and a daily report.
Envelope per trace_id, schema validation and end-to-end auditing.

Live demo

Hi! My team lives in copy-paste and I lose leads because support doesn't scale.

Analyzing…

Lead qualified · notified on WhatsApp

Stack

AI

For a tech client (anonymized).

Proposals built by hand — slow, inconsistent and error-prone on critical data.

AI agents with anti-hallucination guard-rails that never alter tax IDs, price or deadline, deterministic validation and a ready-to-send document.

Premium product engineering. No AI label — here the proof is craft: offline-first, tested, fast.

My own product — production-ready.

Engage and retain through a 21-day challenge, even offline.

An offline-first, gamified PWA with an offline sync queue, its own design system and 700+ tests.

Sites with product craft, performance and real perceived value — some already published.

+ a tech client (whitelabel)