← The SHaDS™ checker
SHaDS™ · for organisations

A governance standard for AI-written work

The free checker warns you what to check. A governance instance goes further — your own version-controlled checks, used the same way across every team, setting what your organisation won't let through.

🔒 Disconnected by design — every check runs inside your own AI. Nothing you paste leaves your environment.

The problem it governs

Bids, board papers, policy drafts, marketing copy, HR letters, clinical and legal summaries — more and more of them are written with AI. Each one can quietly carry the five SHaDS defences: smoothed-over doubt, confident claims with no ground, performed authority, drift off the actual question, and agreement in place of reasoning. Most organisations have no shared, repeatable way to catch them before they go out the door.

SHaDS™ gives that a name and a check. For a single person, the free tool is enough. For an organisation, the check itself has to be governed — owned, versioned, and consistent across teams. That is the difference between a handy prompt and a standard.

What an organisation instance is

🗂️ Your own library

A set of SHaDS checks tuned to your material — the bid response, the board paper, the clinical note, the customer email — not one generic prompt.

🔖 Revision-controlled

Every prompt carries a version. You always know which version produced a given check, and nothing changes under you without a new number and a changelog.

🔁 Learns from feedback

When your people flag a miss or a false alarm, that feedback feeds the next revision. The library gets sharper on your work, on the record.

🔐 Gated & private

Your library is yours — held privately, shared only with the people you name. The checks run inside your own AI; the material never leaves your environment.

Three ways to use it

01 · SELFRun it yourselfAny team member, on their own chat or document, in two minutes. This is the free checker — always open to everyone.
02 · TEAMA shared standardThe same versioned checks across marketing, bid, legal, HR — one governance language, so a green light means the same thing everywhere.
03 · ASSUREDReviewed with PaulA documented SHaDS governance pass on your real material, the library maintained against your feedback, and training for your people.

How it works

  1. We scope your material. Which document and conversation types matter most — and what "good" looks like for each.
  2. I build your library. A first set of SHaDS checks, each versioned, each disconnected — designed to run inside whichever AI your organisation already uses.
  3. Your team uses it and feeds back. Misses and false alarms are logged; each becomes an input to the next revision.
  4. The library is maintained. New versions are issued with a changelog, so you always know what changed and why — and can roll back.

The one rule that makes it trustworthy

A tool built to catch made-up facts must never make up its own verification. Every checkable claim is marked ✓ proven (with a real, named source), ✗ disproven (with the correction), or ⚠ unverifiable — said plainly. It shows its working, and admits what it cannot confirm.

Ask for an instance

Tell me a little about your organisation and the work you'd want checked. I'll show you what a private SHaDS library for your teams would look like — starting from a live example.

Email hello@paulroebuck.co.uk →

⚠ Beta & experimental

SHaDS™ is an unproven, untested method, offered as-is with no warranty and no liability — a governance aid, not fact, not a diagnosis, not advice. Never rely on it for medical, legal, financial, clinical or other consequential decisions. Not for therapeutic, clinical or emotionally sensitive material.