Questions we are asked.
The questions serious buyers put to us before they will talk, answered plainly. Where the honest answer is that we do not publish something, that is what it says.
About the product
What is Quorum?
Quorum is a platform for running governed decision simulations. You bring a bounded, high-stakes matter; the platform runs it as a governed process in which the brief is pre-registered and challenged before any evidence is gathered, an independent system examines the evidence and cross-examines itself, every material claim is traced to a source and graded for confidence, and a named person in your organisation signs off. What you receive is not a report to take on faith. It is an evidence base with an audit trail.
How is Quorum different from an AI research tool?
An AI research tool optimises the answer. Quorum establishes the answer’s evidential status. A citation beside a sentence proves a source exists; verification proves that the source supports the specific claim, and that two sources agreeing are not the same source arriving twice. Quorum also reports what it could not establish, the part that is structurally invisible in a standard AI report, and records who approved what, on which evidence.
What do I actually receive?
A written simulation answering the pre-registered question, and underneath it the evidence base. That means a claim-level record showing, for every material claim, its sources, the level of corroboration and an explicit confidence grade. It also includes the declared-absence section, covering what could not be established, where the record is silent, and the conclusions the simulation declined to draw. Then the record of disagreement inside the system and how it was resolved, and the audit trail: original brief, every revision with its stated reason, and the sign-off.
How long does a simulation take?
It depends on the matter, and we do not publish a standard timeline because publishing one would be misleading. Scope, the state of the public record, and how much of your own material is in play each change the answer materially. What you do get is a delivery window set in writing before the simulation begins, alongside the agreed scope. If a matter cannot be answered to standard inside the window you need, the platform says so at intake rather than after.
What is Compass, and do I need it to run a Quorum simulation?
Compass is a separate product: a standing intelligence platform that maintains a continuous picture of your domain and of your own knowledge base: what is emerging, what your portfolio contains, and where you sit relative to the field. You do not need Compass to run a Quorum simulation, and you do not need Quorum to use Compass. They run over the same evidence layer, so holding both compounds: Compass surfaces which matters are worth examining, and a Quorum simulation’s verified findings return to the standing picture.
About the method
What does “independent” mean here?
It means two sources, or two perspectives, are not derived from one another and did not both inherit the same upstream material. Independence is assessed per claim rather than inferred from agreement, because several systems saying the same thing tells you very little if they read the same document or share the same lineage. Where independence cannot be established, the corroboration is recorded at a lower level rather than counted at full weight.
What happens when the system disagrees?
Disagreement is treated as a finding, not as noise to be averaged away. It is escalated to structured challenge, where the competing positions are tested against the evidence and resolved by an accountable decision that is recorded with its reasoning. Where the disagreement cannot be resolved on the available evidence, it stays in the output as an open question with both positions set out, because that is the true state of the evidence, and you are entitled to know it.
What if the evidence doesn’t exist?
Then the simulation reports that, and it counts as a result. “We could not establish this” is a finding, frequently the most valuable one in the simulation, and it appears in the output rather than being written smoothly into prose. A simulation that returns a well-evidenced negative, or a clear map of where the record is silent, has done its job.
How do you stop the system inventing sources?
Every material claim carries its own evidence record, and a source only counts once it has been checked both as existing and as supporting the specific claim attached to it. The second check is the one most systems skip. Whoever gathers the evidence does not grade it. That separation of duties is structural, the same discipline that makes audit and editorial work trustworthy. Claims that fail verification are removed or downgraded rather than softened into vaguer language, and the downgrade is visible in the record.
Can I see the working?
Yes, and that is the point of the product. Any conclusion traces backwards to the claims it rests on, each claim to its sources and its grade, and the whole simulation forwards from the pre-registered question through every revision. Nothing in the method requires you to take our word for anything.
About governance and control
Who signs off?
Your own people. Sign-off sits at four defined gates rather than as a formality at the end, and the platform does not release unreviewed work. You nominate a simulation owner who approves the pre-registered brief and any revision to it, and the reviewer who rules where the system cannot resolve a disagreement. Serinyx does not rule on the substance of your simulation.
Can I change the brief mid-simulation?
Yes, with a record. Revisions are permitted and versioned: the original brief, the change, the reason and the date remain permanently in the audit trail. What the platform will not do is quietly re-shape the question to fit what the evidence turned out to support. That is precisely the failure pre-registration exists to prevent, which is why a revision is a visible event rather than an edit.
What is pre-registration, and why should I care?
Pre-registration means the question, the scope, what would count as an answer and what would count as disconfirming evidence are all fixed in writing before any evidence is gathered. It matters because most research, human and machine alike, settles what it was asked while answering it, and the output reads as coherent either way, so nothing on the page distinguishes a finding from a rationalisation. Lock the brief first and the difference becomes visible afterwards. A brief that cannot fail is refused rather than answered.
What if I disagree with a finding?
Challenge it against the evidence record; that is what the record is for. If the challenge identifies evidence the simulation did not reach, or a grading that cannot be defended, the simulation is corrected and the correction is logged with its reason. If it does not, your disagreement can be recorded alongside the finding. A confidence grade does not move because a conclusion is unwelcome.
About scope
How is the scope of a simulation set?
At the mandate stage, in writing, before any research work begins. The matter is framed, challenged, and broken into discrete answerable questions, and your named owner approves that structure. Nothing downstream starts until it clears.
Can the scope run away?
No. The scope is fixed at the mandate stage and cannot expand without your written approval. If the matter turns out to be larger than the mandate covers, you hear that during the simulation, with the choice to narrow it or extend it. Not afterwards.
About data and security
Who can see my data?
Your material is available to your own people working in your tenant, and to no other organisation. Access is scoped to your organisation, and the audit trail records what material was used and where it appeared in the output. Your simulation, your brief and your evidence base belong to you.
Is my data used to improve the system for other clients?
No. Client data is isolated per client. It is not used to train models and it does not enrich another client’s evidence base. Cross-client pollination does not occur, and that is a design constraint rather than a policy we could quietly relax.
Where is data held?
Data residency is configurable and agreed before a licence begins. For public-sector buyers and for organisations under supervision this is a scoping question rather than an afterthought, and we treat it that way. If you have a residency requirement we cannot meet, we will tell you at the outset instead of discovering it in procurement.
Getting started
How do I get started?
Start with the matter. We hold a scoping conversation, then the platform runs an adversarial intake in which the brief itself is tested: what would count as an answer, what would count as disconfirming evidence, and whether the question is capable of failing at all. You then hold a pre-registered brief, a fixed scope and a delivery window. Work begins when your simulation owner approves them.
What do you need from me?
The matter, the decision it feeds, and a named owner who can approve the brief. Optionally, any internal material you want in scope: prior analysis, portfolio documents, data-room contents, regulatory correspondence. You do not have to provide any of it; a simulation can run entirely on the public record. Where you do provide it, it stays isolated to your organisation.
Can we run a trial simulation?
Yes. A first simulation on a real but bounded matter is the normal way to begin, and it is how most buyers test whether the method holds up on ground they already know well. We would rather you run a small simulation you can check against your own knowledge than a large one you have to trust.