AI PRINCIPLES AND GOVERNANCE
Version 2 · Approved October 13, 2026 · Effective at launch
We assess how organizations stand behind decisions that involve AI. This page sets out where we use AI in our own work, what it may and may not see, who decides, and what you can choose. It is written so you can check us against it.
01At a glance
AI helps us research, draft and organize. It never scores, and it never decides anything about you.
Settings and notes, never the contents of your emails, files or chats. Never used to train AI models.
The Monarc Principal Architect reads every request, assigns every score and signs off every deliverable.
Decline recording, ask what AI touched, or ask us to keep AI off your material.
02Where we use AI
If an activity is not listed here, we do not use AI for it. When we add a use, we update this page first.
| Activity | AI used? | What AI does | Who decides |
|---|---|---|---|
| Requests sent through our website | No | Nothing. Forms carry a security check that blocks automated spam; no AI reads, screens or ranks your request. | The Monarc Principal Architect reads every request. |
| Deciding whether we can help you | No | Nothing. | The Monarc Principal Architect, after a conversation with you. |
| Research and our published writing | Yes | Helps find sources and draft text for Insights and the Brief. | A person checks every source; every figure names its publisher and date. |
| Assessment session notes | Only if you agree | If you agree in writing to a recording, AI drafts evidence notes from the transcript. | The Monarc Principal Architect reviews every note. The transcript is deleted once the notes are accepted. |
| Scoring | No | Nothing. The drafting tools we build have no place to record a score. | The Monarc Principal Architect assigns every score. |
| Reading your AI systems (the AI Diagnostic) | Limited | May sort system settings and inventory entries into categories. It sees settings and metadata only. | A named person confirms every system and every result. |
| Your deliverables | Yes | May help with drafting and layout. | The Monarc Principal Architect signs off; each deliverable says where AI assisted. |
03AI and your personal data
We use Claude, from Anthropic, for drafting and classification, with model training turned off in our account settings. Our automation platform, Make, moves information between our systems and does not apply AI to your data. We name any new provider here before we use it.
What we read about your systems stays in a workspace you own. We keep only the engagement record, a fingerprint that shows whether a delivered record has been changed, and summary counts you agree to share. Session transcripts, where you allow them, are deleted once the notes are accepted.
We do not make any decision about you, or anyone in your organization, using automated processing alone. A person is responsible for every decision we make.
Our Privacy Policy explains how to access, correct or delete your personal information, and Your Privacy Choices lets you act on them.
04Your choices
Sessions run the same without a recording. The Monarc Principal Architect takes notes by hand.
Tell us before the engagement starts and we will do the work without AI assistance. It may take longer; we will tell you how much before you decide.
Each deliverable notes where AI assisted. Ask, and we will walk you through it.
05Our principles
AI may help draft evidence and rationale. It never assigns a score, and no AI-drafted material reaches a client without review by the Monarc Principal Architect.
How we hold to it: the drafting tools we build are designed with no place to record a score, and review by a person is a required step, not an option.
We do not process client material with third-party AI services until an agreement covering that use is in place.
How we hold to it: a data processing agreement is a gate before any client material is used with an AI service.
What we design for clients is built in systems they own. Monarc does not host client decision data.
How we hold to it: registries and records are delivered into the client's own workspace.
Clients are told where AI assisted in preparing their deliverables.
How we hold to it: deliverables note AI assistance.
Statistics name their publisher and date. Interpretation is labelled as ours.
How we hold to it: an evidence register records the source behind every figure we publish.
Every scoring method is tested at its limits and against hand-worked cases before it reaches a client, and every change is versioned and recorded.
How we hold to it: material changes to our method carry a written validation record.
We are not affiliated with any regulator, insurer or AI vendor, and no vendor or insurer pays us to influence a score.
How we hold to it: see our Code of Conduct.
06Governance
The Monarc Principal Architect is accountable for this page, for every use of AI it describes, and for every score Monarc issues.
We review this page at least once a year, and before we adopt a new AI tool or change how we use one. Each version is dated. If a change affects how we handle your material, we tell active clients by email before it applies to them.
If you believe we have not met what this page says, tell us through our contact page. Every message is read personally.