Designed & built with Claude.
The footer says this site was designed and built with Claude, and that its updates ship by an AI agent. That is true — and for a firm that advises on AI governance, you deserve the full version rather than a tagline. Here it is.
The short version
This website was designed and built with Claude — Anthropic’s AI — and its updates are shipped by an AI agent working under my direction. I decide what gets published, I check it, and I own the result. This page sets out exactly what that means, and where the lines are.
What the AI does, and what I do
The division of labour is deliberate — and it is the same one I teach.
| The work | Who does it |
|---|---|
| Building and styling the site | The AI agent (Claude Code) — on my instructions |
| Drafting and updating the words | Drafted by AI, then edited by me |
| Shipping a change — build, version control, deploy | The AI agent — once I approve it |
| Deciding what is worth publishing | Me |
| Checking the facts, above all the law | Me — against primary sources |
| Professional responsibility for everything here | Me |
The method behind it
I work to a four-part framework — my adaptation of the AI-fluency model (Anthropic / Dakan / Feller) — and this site is built with it, not merely described by it:
- Delegation — choosing what to hand the AI, and what stays with me. Judgement, positioning and the final call stay human.
- Direction — briefing the AI clearly enough to act on a goal. Much of that brief is built into the setup itself, not retyped each time.
- Evaluation — reading every output critically before it goes live: facts, logic, tone, fit.
- Ownership — my name on the result. “The AI did it” is never an excuse.
The longer version is itself a guide: delegate the goal, not the task explains how I direct an agent, and the files that run an AI practice shows the plain files that make it repeatable. The same architecture, scaled to an organisation, is the question of who owns AI governance.
Accuracy, and the law in particular
A confident-sounding draft is not the same as a correct one. So anything factual here — above all the dates and obligations under the EU AI Act and the GDPR — is checked against primary sources before it is published, and re-checked when the law moves. The machine-readable summary some AI assistants read (llms.txt) is held to the same standard: a stale fact that an AI then repeats is precisely the failure I warn clients about, so I will not knowingly ship one. If you spot an error, please tell me and I will correct it.
Where AI is not used
- No client data is used to build or run these websites.
- Client engagements carry their own governance and confidentiality rules; nothing here describes how I handle a client’s information.
- For how personal data is processed — including which AI tools are involved and the safeguards in place — see the privacy policy.
Anchored in the frameworks I advise on
Transparency about AI is not my invention; it runs through the rules I help clients meet. The EU AI Act builds in transparency duties (Article 50): telling people when they are dealing with an AI, and flagging AI-generated content. Tellingly, its duty to flag AI-generated text put before the public falls away precisely where a human has reviewed the content and holds editorial responsibility for it (Article 50(4)) — which is exactly the line I hold here. So this page is not a legal box to tick; it is me being explicit even where human ownership already answers the rule.
The same principle sits at the centre of the others. Transparency is the first principle of the GDPR, which also gives people the right to know about automated decisions; and the governance frameworks I work with — the OECD AI Principles, ISO/IEC 42001 and the NIST AI Risk Management Framework — all put transparency and human accountability at their core. I hold myself to the spirit of all of them.
Why I publish this
Because it is the standard I would ask of anyone else. A consultancy that advises on AI literacy and governance should be open about its own use of AI — and able to show, not merely assert, that a human stays accountable for the output. You are looking at the method in production.