The rules of the language
This is the part a general-purpose model cannot do consistently. Boards have a vocabulary, and the vocabulary is what creates the record.
“Noted” creates nothing. “Approved” creates a decision. “Agreed in principle” creates a decision with a condition. “The CEO undertook to” creates an action with an owner. A chair knows all this instinctively. Quorums knows it as written rules, and applies them the same way to every document.
The difference matters because a chatbot also knows these words. What it does not have is a rule set it is bound by. Ask it twice and you may get two different answers about whether “the board noted the recommendation” was a decision. Ask Quorums twice and you get the same answer, and you can read the rule that produced it.
Why this cannot be a feature of a chatbot
A general-purpose model has a world model: a rough, impressive sense of what a decision or a risk usually is. An ontology is domain specific: it says what kinds of thing exist, what they are called, how they relate, and it lets the system reason over those relationships rather than match words on a page. It is a shared, unambiguous layer of meaning, not a pile of text. An ontology is a commitment.
That distinction is not academic. Because the five kinds of record are fixed, we can run every change to the system against a frozen set of board documents and count the errors. When a newer model asserted fourteen things as record that were not in the documents, the ontology is what let us count to fourteen. Without a fixed definition of a decision there is nothing to count against, and “it seemed pretty good” is the only quality measure available.
What it does for you
A decision in the March pack and a decision in the September pack are the same kind of thing and can be laid side by side. A metric from five packs is five comparable figures, not five paragraphs. An action can be followed from the moment someone undertook it to the moment it was done, or the moment it quietly vanished. Across every board you sit on, the same question gets the same kind of answer.
And every one of those things points at the sentence it came from. The ontology tells Quorums what to look for. The evidence tells you it found it.
The short version
Boards never had a written grammar. Quorums has one. Five kinds of record, one kind of observation, the rules of the language that create them, and a sentence of evidence behind each. That is a board ontology, and it is the reason a purpose-built system beats a brilliant generalist in the room.