Skip to main content

    Interactive assessment

    The STOP Protocol

    A four-step discipline for reviewing AI-generated outputs before acting on them.

    Use this reference card before sharing, publishing or making decisions based on AI-assisted work.

    Dimensions
    Stop, Test, Optimise, Proceed or Pause again

    What it covers

    • Stop
    • Test
    • Optimise
    • Proceed or Pause again

    How this interactive tool works

    This page provides the documented overview before the interactive experience is loaded. It lets you read the purpose, the material covered and the way the result is presented without entering answers or sending a request. When you choose to continue, the interactive version uses the same material shown here. It then presents the route's existing prompts, controls and completion behaviour.

    The interactive experience remains responsible for its own input, score and lead-capture step. Where the route requests a first name and work email, its existing form handles validation and submission. This overview does not replace those steps or submit a lead.

    What you will work through

    These are the sections and prompts already used by this route. They describe the scope of the interactive experience and the information it asks you to consider before it creates a result, printable card or saved browser document.

    Stop: pause before you act

    Build a deliberate pause into the workflow before accepting, sharing or acting on AI content.

    • Whether the output has received genuine critical attention or only looks credible
    • What happens if the output is wrong, misleading or biased
    • Whether an assumption that AI would not make that mistake is justified

    Test: verify accuracy and reliability

    Check the output against knowledge, additional sources or domain expertise. The degree of scrutiny should match the stakes of the decision.

    • Whether facts, figures and citations are accurate and verifiable
    • Whether common errors such as confabulation, statistical misinterpretation or cultural bias are present
    • Whether the output could be attributed to you after subject matter review

    Optimise: refine before it leaves your hands

    Treat AI output as a starting point and add the human judgement, context and nuance the tool cannot supply.

    • Whether the output reflects organisational context, constraints and priorities
    • What human judgement, relational insight or professional experience needs adding
    • Whether the tone, framing and emphasis fit the audience and purpose

    Proceed or pause again

    Act only after stopping, testing and optimising. Where material doubts remain, pause and seek further input.

    • Whether you are confident enough to put your name to the output
    • Whether you can explain the reasoning behind a consequential decision
    • Whether a colleague, domain expert or additional source is needed

    What happens after you complete it

    This is a printable reference card, not a scored assessment. The completed card distinguishes low-stakes work, where a quick sense-check may be appropriate, from medium-stakes work needing a full STOP review and high-stakes work requiring a full review plus expert verification.