Skip to content

Why not just use Claude?

Quorums runs on Claude. So, the question is fair: why pay extra for this system when the model underneath is a subscription away?

Because the model is not the product. Here is what is.

The Tuesday before the board

Three docs arrive this week, totalling hundreds of pages. Somewhere in the second file is a risk you think came up last autumn and a margin figure that does not match the one you saw in the spring.

Put that bundle into a general-purpose assistant and ask for the decisions. You will get a confident, fluent and well-written list in seconds. I have done it. It is impressive … right up to the moment you say out loud “I thought we agreed that in March” and the CEO says “we didn’t”, and she is right.

That is the failure mode. Not unhelpful. Fluent, plausible and wrong, with no way of telling which. In a boardroom that costs the one thing a board member seeks to maintain: their credibility.

“What did the board decide about the pricing change, and what happened next?”

A general-purpose assistant

The board approved a change to the pricing model at its June meeting, following a recommendation from the CFO. Management was asked to implement the new structure and report back on customer impact. Follow-up appears to have been planned for a subsequent meeting.

Sounds right. No date, no wording, no owner, no way to check.

Quorums

Decision, 14 June 2026

Agreed in principle, not approved. Conditional on the revised forecast.

“The Board agreed in principle to the revised pricing structure, subject to the updated FY forecast being tabled in September.”

Action, from the same item

Table the updated forecast. No owner recorded. Still open at the September pack.

Open question, 14 June

Customer impact of the change. Not answered in any later pack.

Synthetic board (Northwind). The gold mark is a watch flag: the point worth returning to.

Boards never had a grammar. So we wrote one.

Accountants produce ledgers, journals and statements, and every number in a set of accounts is one of those things, with the same rules applied whoever prepared it. Boards never really had that before. They produce decisions, actions, risks, open questions and metrics, meeting after meeting, and leave them scattered across a pack, a set of minutes and somebody’s memory.

Quorums is built on a board ontology: a fixed, written account of the five kinds of thing a board cares about, what each is made of, how they relate, and which words in board language bring them into existence. “Noted” creates nothing. “Approved” creates a decision. “Agreed in principle, subject to the revised forecast” creates a decision with a condition, and the condition is a field, not a footnote. “I’ll circulate the numbers by Friday” creates an action with an owner, even though nobody said the word.

A chatbot knows all these words. What it does not have is a rule set it is bound by. Ask it twice whether “the board noted the recommendation” was a decision and you may get two answers. Ask Quorums twice and you get one, and you can read the rule that produced it. A general model has a world model. 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.

Numbers you can put side by side

In chat, a metric is a sentence. In Quorums it is a value, a unit, a period and the place in the document it came from. Ask how gross margin has moved and you get five figures from five packs, each comparable with the last. A number that cannot be matched to its source is marked unverified, not smoothed over.

A memory that persists

A chat is a user thread. A board memory spans years. The questions that matter usually fall between meetings: what did we say last time, did the action happen, which risk has sat unmoved for a year. Quorums holds the archive for every board you sit on and lets you interrogate it across time, from one seat, each board sealed off from the others.

It shows its working, or it says no

Every decision, risk and open question Quorums surfaces points at the sentence in the source it came from. When it cannot find the sentence, it tells you so. It would rather say “I can’t find that” than guess.

We enforce this with a measurement, not a slogan. Earlier this year we trialled a newer, more capable model. On our fixed test corpus it asserted fourteen things as matters of record that were not in the documents. We did not switch. The ontology is what let us count to fourteen; without a fixed definition of a decision, “it seemed pretty good” is the only quality measure left. Ask your chatbot vendor for its invention rate on a board pack. We know ours.

Where did the pack go?

Would you tell your CEO exactly where the board pack went when you dropped it into your personal assistant? Quorums is the answer you can give a board with a straight face: every board isolated at the database level, nothing used to train anything, a published list of processors with a contract behind each. A chatbot is a tool you use quietly. Quorums is one you can declare.

What it will not do

Record meetings, produce the statutory record, assess directors, or tell the board what to conclude. A chatbot will happily do all four. A chair who lets a tool grade the room has changed the room.

The short version

Claude is brilliant and has no memory of your board. Quorums is the part that knows what a board means, remembers every one you sit on, shows you the evidence, and refuses to make things up. You do not choose between them. You put Quorums on top.

Worth a conversation?

The first conversation is exactly that: a conversation about your boards, not a sales call.

Request a conversation