The problem
The project grows. Knowledge scatters.
The more complex the product, the more its truth fragments: across archives, spreadsheets, email threads and people's memory. These are universal problems of make-to-order production — and none of them is solved by "more discipline".
Scattered documentation
Specifications, reports, plans and certificates live in different formats and archives. Finding the right version is a job in itself.
Planning on spreadsheets
The multi-year production plan — spaces, timings, margins — lives in sheets only their author knows how to read.
Decisions without a trace
An authorisation, a test outcome, a signature: when you need to reconstruct the why, the source is nowhere to be found.
The time nobody sees
The project's lead time — from first plate to delivery — is never visible in full to anyone.
Critical knowledge in heads
How to build a test plan, what to check at goods-in: knowledge of a few key people, not of the system.
The vision
The "Piano Nave 4D" — a 4D ship plan
Picture the project as a four-dimensional object: the three dimensions of its physical breakdown — zone, deck, room, system, component — and the fourth, programme time, from design to launch, from sea trials to delivery, all the way to refit. The 4D Ship Plan is the digital spine holding it all together: one structure that every module, document and decision plugs into.
Every piece of data has an address
Every document, test and decision is anchored to a point in the product breakdown — zone, deck, room, system, component — and to a moment of the programme. No more "which folder was it in?".
AI reads the yard's documents
Language models extract data from specifications, reports and archives, and fill the drafts. Every field carries a source-fidelity traffic light — green, yellow, red — with the source document one click away.
Decisions stay deterministic
States, authorisations and progress are state machines generated from specifications and covered by tests. The backbone of the system is not the model: it is code you can audit.
The future can be simulated
What-if scenarios on space saturation, volume curves and margins: the planner sets the strategy, a constraint solver does the maths.
What plugs into the spine today
The principle
The LLM proposes, the deterministic engine decides.
In production processes a documentation error has legal, contractual and economic consequences. That is why we confine AI to what it excels at — reading documents, extracting data, composing drafts with evidence — and entrust every decision to deterministic rules generated from specifications and covered by tests: traffic lights, states, formulas, constraints. The result is an auditable system, where you can always answer the question "why is this status red?".
In the field
The 4D Ship Plan comes to life in a multi-year AI-for-production programme at a leading Italian superyacht shipyard — dedicated cloud platform, corporate SSO and modules in progressive rollout.
Methodology
From specification to code, from code to audit.
Canonical specification
Every process is described by a canonical model from which we generate types, states, checks and APIs. The system's behaviour is defined before the code — and verifiable against the specification.
The AI reads and proposes
Language models read technical documents, emails and archives, and produce structured, schema-validated drafts — with evidence and precise references to the sources.
The engine decides, the expert validates
Traffic lights, states and constraints — up to constraint-solver optimisation for planning — are computed by tested deterministic rules. The expert steps in on the doubtful cases flagged by the system.
Where does your project lose time?
We start from an assessment on a real case: together we measure the skilled time spent on compilation and checks, and we show you — on your domain — what the 4D spine can take over.
Book an assessment →