The model, in person
The twin can be visited, not only queried. Four avatars give it a face and a voice. All of them read from the model and only from the model.
The voice of the position. It explains the JOB, its tasks and its RACI, from inside the desk.
The voice of the flow. It explains what comes before, what comes after, and why.
The host of the interview demo. It questions the other avatars along a script, takes the audience's questions and closes with a summary.
It plays an external party: a client, a supplier, a candidate. A person trains by talking to it. It always declares that it plays a role, and it works on a defined scenario, never on real client data.
A walk through the twin is a sequence of stops. Each stop carries one concept, a short mandatory part of about fifteen seconds and an optional deeper part. The author writes the structure and the intent. The AI generates the words at the moment, in the visitor's language.
An invisible role, the DIRECTOR, holds the state of the scene. It decides who speaks, sends each question to the avatar with the right point of view, and avoids repetitions.
The look of an avatar is separate from its logic. Neutral appearances serve school and onboarding. A specific likeness can be used for a prospect demo. A robot appearance is reserved for desks where an agent executes the work.
In training, the guide avatars teach the real company. The course is generated from the model and stays current because it is the model. The same avatars that assist a person at work are the ones that trained that person.
When a critical person approaches retirement, their knowledge is captured in structured form: interviews, solved cases, rules of judgment. It becomes an avatar linked to their tasks. The successor can ask it questions after the person has left.