Five views of one model
The Siege Ontology Framework presents one model of the business through five views, which we call cuts. Each cut answers one question for one audience. Every cut reads the same model, and nothing is copied into a cut, so no two cuts can disagree.
Cut | The question it answers | What it gives the business |
|---|---|---|
Altitudes | What is the business made of? | The business at different heights, from the whole company down to one team's working parts. The owner reads it from the top, a team lead at their own level. Same model, at the height each person needs. |
Perspectives | What happens in this area of work? | One area of work, such as quoting, fulfillment or billing: what it reaches, what may be done there, and by whom. Each team sees its own work, and only its own work. |
Axioms | What must always be true? | The business's own rules, in the owner's words: "nothing ships before the customer approves it", "a rush order needs a manager's approval". These are rules the business states, never engineering habits put in their place. |
Governance | Who decides, and on what authority? | Who may do what, which decisions the business has made (with who made them and when), every change to the model, and the boundaries every agent works inside. |
Analyses | What does the business need to know back? | Every recurring answer the business relies on: what it draws on, who it is for, and how often. A report is designed into the model, not bolted on afterwards. |
Why the cuts are kept separate
In most software these five concerns end up tangled. A business rule gets buried in a screen's code, permissions sit in a configuration nobody reads, reports are built afterwards in another tool, and the process lives in someone's head. Each concern is maintained on its own, so each drifts on its own.
Because the five cuts read one model, a change made in one is present in all of them:
A rule changed as an axiom is the rule every screen applies, every agent obeys and every report assumes.
Widening what a kind of person may do, in governance, widens it on every perspective where they work, and the change says so before it is made.
A new analysis draws on facts the model already holds, so its numbers mean what the rest of the system means.
Each audience reads its own cut without the others: the owner reviews axioms and governance, a team lead their perspective, an adviser any of the five. All of them read the same model.
The thesis file
The whole of a client's Ontology is held in one file: the thesis file. One file, one numbered revision, the complete model.
It can be audited. Every revision is kept with who changed it, when and why, and any revision can be exported whole, so two can be compared with standard tools.
It is the client's. The client receives its own copy of the model in its own repository, in an open W3C standard (RDF/Turtle) that standard tools read, so the description of the business never depends on Siege to be read.
It is light enough to hand to an AI. An agent can be given exactly the part of the model its task needs, at the moment it needs it.
An agent works from the model's exact rules, not from a summary of them. Exact text alone does not stop a language model misreading it, so the other half of the control is that every action an agent takes is checked against the same rules before it takes effect, and any text it drafts reaches a customer only when a person sends it. An action it holds no key for is refused at that check.
Every revision is published to the client's repository as it lands, so the client's copy never trails the live model.