Skip to main content
WHAT IT IS

The working A prototype: use its controls to test the build question.

WHY IT MATTERS

Determine whether a human reviewer can actually change or stop an AI-assisted decision.

RN'S ROLE

RN conceived the question, structured the evidence boundaries, designed the experience, implemented the prototype and documented its limits.

BOUNDARY

The generated protocol is a design aid, not a substitute for legal, clinical, safety, or domain-specific review requirements.

HOW TO USE THIS PAGE

Work through the controls with a situation of your own rather than the sample values. The tool responds to what you put in, so the useful output comes from real input, and everything is processed in your browser as you go. Any sample content you find already loaded is there to show the shape of a filled-in state, and you can clear it and start again at any point.

WHERE TO GO NEXT

The build overview sets out the question, the purpose, the role and the evidence status in one place, and the public record behind it states what changed, what another person can reuse and what supports the system. If you want to judge how far this build should be trusted, the record is the page to read, not this one.

BUILD 001-A
ON VIEW

Human Review Design Framework

AI is often described as “human reviewed.” This tool helps you check what that really means. Can the person reviewing the AI see enough, act in time, disagree, and stop a bad outcome?

Who can use it

You do not need to know how AI works, know the law, or work in technology. If AI helps make, recommend, rank, approve, flag, draft, or influence a decision that affects people, this tool can help you inspect the human review around it.

Start here

Use the 60-second Quick Check for the simplest path. It asks what actually happens in the workflow and tells you what looks strong, what looks weak, and what to fix first. If you need more detail, open Design the Control.

Why these questions

The deeper framework is informed by recurring human-oversight concepts in NIST AI guidance, the EU AI Act, and UK ICO guidance. The tool translates those ideas into ordinary language first, while keeping the more precise concepts available underneath.

No AI expertise needed

Pick the answer that is closest to what happens today. If you are not sure, that uncertainty is useful information too. You will get a simple diagnosis and the three things to fix first. Your answers stay in this browser.

01What is the AI helping with?Choose the closest example. You do not need to use technical terms.
02Who checks the AI?A human reviewer only helps if they know enough and have enough time to think.
03When do they check it?A review that happens after the important action may be too late.
04What can the reviewer see?People cannot challenge an AI answer if the answer is all they are allowed to see.
05Can they change what happens?A reviewer without real authority may only create the appearance of oversight.
06Can you prove the review happened—and learn from it?A useful review process leaves a record and gets better when problems appear.
Plain language first · deeper detail optional · no account
What the result means

A person being present is not the same as a person having control.

Your result is a design check, not a legal grade. It shows where human review looks strong, where it looks weak, and what to fix first. A higher score does not prove the system is safe or legally compliant.

If the tool finds a serious gap: ask what has to change so a real person can understand the situation and affect the outcome before harm happens.

If you are working through a related systems problem and want another set of eyes on the architecture, contact RN.