How it works

Same inputs, same findings, always.

Deterministic is a strong word, so here is exactly what it buys you — and, just as importantly, what it does not.

What determinism means here

Reproducible,
or it is not evidence.

An examiner asks why a gap was or was not raised eight months ago. If the answer depends on which day the report ran, it is not an answer. The engine is built so that the same register, the same rule pack and the same as-of date always produce the same findings.

The date is always explicit

Every run is evaluated as of a stated date. Nothing depends on when you happened to press the button, so a finding from March can be reproduced in September and still read the same.

Findings are stably identified

The same underlying condition keeps the same identity between runs. A gap does not vanish and return as something new because a row moved — which is what makes tracking a remediation possible at all.

Rules are data, with versions

Detection logic encodes no state’s requirements. Rules live in a versioned pack, so any finding traces back to the exact pack that produced it.

Every cell traces to a source

Each rule entry carries its citation. A finding you cannot follow back to an authority is not evidence, and will not survive the conversation it exists for.

What is not determined

Faithful to the pack.
Not to the law.

This is the distinction most compliance software blurs, and the one a regulated buyer should press hardest on. Determinism guarantees consistency. It guarantees nothing about correctness — and we would rather draw that line ourselves than have it drawn for us during an exam.

AI drafts rule research and remediation language for human verification. It never determines whether a gap exists; that path is deterministic code.

Whether a rule is correct

The engine is faithful to the pack. Whether the pack is faithful to the statute is a human question, answered by a person reading the source and putting their name against it.

Whether you are compliant

Not a claim this product makes, on any surface, ever. It reports that no licence in your register covers an activity at a location. What that means for your position is a judgment with professional consequences, and it belongs to you and your counsel.

What the register does not contain

Detection runs against what you supply. A location nobody recorded produces no findings — which is why intake reports what it rejected and why, instead of quietly moving on.

Silence is made visible

An uncovered activity
is a question, not a pass.

Where an activity falls outside the pack’s declared coverage, the result is no_rule — an open research task carried into the report. It is never reported as the absence of a gap. The most dangerous output a compliance tool can produce is a clean page that actually means “nobody looked.”

AZCAFLGAILNVNYTX

Rule pack 2026.08.1-draft, effective 2026-08-01 · 13 entries · 13 drafted and not yet verified by a person. Read from the pack at build time, not typed into this page.

Rejections are reported, not swallowed

Intake states what it could not read and why — a malformed date, an unknown activity code, a location with no matching entity. A silent import is how a register comes to disagree with reality without anyone noticing.

The output is a worklist

Each finding names the location, the activity, the date it was evaluated as of, and the rule that produced it. Written to be handed to whoever has to fix it — and read later by whoever asks why it was raised.

If this is your problem

Tell us where the register stops matching the floor.

What states you hold, what your locations actually do, and which licences you would check by hand before an exam.

How to reach us