AI-NATIVE DEVELOPMENT
FORTIFY · TEST
CRAFT
Your team is already using AI to write code. This is the method for doing it with structure — before the deadlines, the insurers, and the category leaders decide it for you.
Three incidents. One root cause.
Orders lost
A six-hour outage that cost roughly 6.3 million orders.
Exfiltrated
A 1.1-terabyte data exfiltration through a typosquatted AI package that ran unmonitored for months.
Prod database
An agent that, given ambiguous instructions during a production freeze, deleted a production database — then fabricated records of the recovery.
Different companies, different failures, one thing in common. And it isn't AI.
These three incidents have something in common, and it isn't AI. It's the absence of method around AI.
This has happened four times before.
Every time a powerful new capability hits engineering, adoption outruns method — incidents rise, vendors sell tools, and the tools never close the gap. Then a handful of organisations name a method, and the ones who adopt it early become the category leaders.
Four times. Same shape. Same outcome. This is the fifth.
The deadlines are already on the calendar.
This isn't a marketing urgency. Three forces are converging on dates that are already set — and two of them are hard:
Insurance
Carriers began excluding AI claims where you can't show documented controls. No governance evidence, no coverage.
EU AI Act
High-risk obligations become enforceable — Article 12 logging is architectural, not retrofittable.
Lands on you
It falls on whoever can't produce evidence. CRAFT produces that evidence as a byproduct of its practices.
The exclusion is not ‘we won't cover AI.' The exclusion is ‘we won't cover AI without governance.'
Five practices. Five verbs.
One non-negotiable order.
Not five nouns you can buy — five things your org has to DO, every day.
Fourteen chapters. Three parts.
- 01 Vibe Coding Is a Business Risk
Forty teams, forty architectures no one designed. - 02 The Anatomy of an AI Failure
How AI velocity outran human-pace governance. - 03 Why Methods Beat Tools
Lean, Agile, DevOps, DORA — and now. - 04 The Window
The insurance, regulatory, and liability deadlines already set.
- 05 The CRAFT Framework
Five verbs, one dependency chain. - 06 Curate
Context as a first-class engineering asset. - 07 Refine
The PRP, and no code without a validated spec. - 08 Architect
Make the right thing structurally easier than the wrong thing. - 09 Fortify
SBOMs, AIBOMs, and defence at AI velocity. - 10 Test
Close the loop with gates AI can't slip past. - 11 How the Practices Interact
Why the five only work as a chain.
- 12 Diagnose Your Org
Score yourself from Ad-Hoc to AI-Native. - 13 The 18-Month Transformation Path
Foundation, Expansion, Independence. - 14 The Category
Adopt the method that defines 2030, or follow the one that does.
From Ad-Hoc to AI-Native, in three phases.
The transformation isn't a workshop or a quarter — it's eighteen months of structured work, and the book gives you the whole map.
Foundation
Stand up context, specs, and the first patterns.
Expansion
Roll the five practices across every team.
Independence
The method runs without its champion. AI-Native.
The methodology you adopt now will shape the methodology your industry adopts in 2030.
Not theory. Forged in production.
Every claim has receipts — Cortex, Veracode, GitClear, Snyk, DX, DORA, and the named 2026 incidents. The method wasn't designed in a boardroom; it was forged building a complete production software ecosystem solo + AI. It wasn't written to sell. It was written because it already worked.
Two authors. One method.
Luis Gonçalves
20+ years in digital product development, 15+ in organisational design. Creator of the SCALEUP, ADAPT, and CRAFT methods. Author of five books on entrepreneurship and agile. Fortune 500 transformation credits. Faculty at Porto Business School.
Jari Laakso
Leadership, coaching, and training across international organisations — teams of 60+ people. Coach to individuals, teams, and organisations, with uncommon attention to detail and relentless learning. The discipline that keeps a method honest in the field.