Not a feature. The foundation.

AI based

The colleague with ten years of experience

AI is at the base of everything in 3D.DET. It reads the whole twin — every structure, task, process, workload, cost, and its history — and answers as the colleague who has been in the company for ten years. Every answer is read from the model, with its evidence. When the model does not contain the answer, it says so.

This cuts onboarding time and errors. Checks carried out in a multinational group with 150 companies worldwide show that 50–60% of help desk and application management (AM) tickets are inquiries. The model answers them, and the cost of help desk and AM falls in proportion.

Building the twin is AI-assisted too: documents, interviews and system extracts are interpreted by AI, then verified and confirmed by people. Humans stay the owners of the truth.

The rule: AI proposes, the human decides

The AI observes, analyzes and proposes. It never writes a fact into the model by itself. Every change passes through a person with the right role, and the model records who decided and on which evidence.

The cognitive engine: Claude

The cognitive engine of 3D.DET is Claude, by Anthropic. Claude reads the model through a standard interface (MCP), within the permissions of the user who asks. It sees only what that user may see.

Three roles

Executor. An AI agent executes a task. The person stays Accountable.
Subject of the organization. Every agent has an identity in the system, defined permissions, a full trace and a human supervisor.
Guardian of consistency. The AI checks the model, finds anomalies and inconsistencies, and reports them.

Seven functions

Navigator

Guides people through the organization: structures, JOBs, processes.

Co-Designer

Supports the design of TO-BE scenarios.

Process Miner

Analyzes the real AS IS processes.

Predictive Analyst

Forecasts trends and anomalies.

Worker 24x7

Executes repetitive tasks.

Memory Keeper

Keeps the organizational memory.

Adaptive Tutor

Personalizes training on the real company.

What the AI does with the twin — progressive agentization — is on the AGENTS page.

In depth

Where AI works in the cycle

Building the AS IS. It reads documents, interviews and system extracts, and proposes structures, JOBs and tasks.
C LEVEL TWIN. It finds the work that leaves no trace in the systems.
Logic mining. It reconstructs the real logic of system transactions.
Gap analysis. It compares the organization with the model of its sector and lists what is missing, what is different, what is in excess.
Initiatives. It detects overlaps and correlations between parallel transformations.
Avatars. It gives the model a voice for demos, training and daily work.

Answers that state their limits

Each fact in the model has a trust class: verified, synchronized, declared, stale or contested. The AI shows it in its answers: "this is verified and 30 days old", "this is declared by the client and 8 months old". When the model has no answer, the AI says so and does not invent one.

Open to other systems

ERP and HR systems, process mining, AI agents of other vendors and partner applications read the model through the same standard interface. They cannot write it. They can submit a proposal with its evidence, and a person decides. Write access to 3D.DET is the right to propose.

AI in the SCHOOL

In teaching, AI generates content from the objective set by the professor, reviews student work against the professor's rubric, tutors the student on the twin, and reads the sequence of decisions the student made. The professor decides the grade.

Why the model makes the AI reliable

A general AI answers from what it has read in general. In 3D.DET the AI answers from the model of that company: its JOBs, its tasks, its processes, its costs. Every sentence can be traced back to the rows of the model it comes from.