Skip to main content
Skip to main content
Back to home
Governance tool

Scenario Builder

An AI risk assessment tool that lets you rehearse the decision before you make it. Describe a proposed use of AI, then test it through both cores of the framework — ALFI for strategy, risk and control, EDIE for dignity, wellbeing and voice — and leave with a readiness score and a governance record you can share.

Delivering responsible AI that people can trust.

Governance portalAcademy courseEthics guide
A team reviewing branching decision scenarios alongside a risk and wellbeing diagram
Scenario readiness
Scored live across 15 responsible-AI checks. Nothing leaves your device unless you ask for an AI review.
Not ready

Overall readiness

0%

0 of 15 checks

ALFI — strategy & risk

0%

Purpose, controls, monitoring

EDIE — people & wellbeing

0%

Dignity, access, voice

Too many core safeguards are missing. Rework the scenario before taking it to anyone.

Still to complete: Scenario title, Purpose, People affected, AI capability, Risks (ALFI), Controls (ALFI), Human impact (EDIE), Safeguards (EDIE), Accountable owner.

Next steps for a educator or practitioner

  • Settle this first: a person decides anything that affects someone materially.
  • Settle this first: accuracy is measured against a known-good sample.
  • Settle this first: the experience meets wcag 2.2 aa and works with assistive technology.
  • Settle this first: the people affected have been asked, not assumed.
  • Trial with one class or caseload before anything wider, and keep the human sign-off visible.
  • Record what pupils, families and your SEND lead told you — that evidence is the voice-of-affected check.
Who is doing this work?
The fifteen checks never change. The wording, the priorities and the AI review are pitched at your role.

ALFI lens for you

Focus on classroom and caseload reality: workload, accuracy against your mark scheme or plan, and who signs off.

EDIE lens for you

Focus on dignity in the room: confidence, SEND and EAL needs, and what a pupil or family is told.

Settle these first

  • A person decides anything that affects someone materially
  • Accuracy is measured against a known-good sample
  • The experience meets WCAG 2.2 AA and works with assistive technology
  • The people affected have been asked, not assumed

Start from a worked example

Author and version your own templates
1. Describe the scenario
Plain UK English. Write it as you would explain it to a governor or trustee.

Produces a draft that a person always reviews and edits.

2. The ALFI lens — strategy & risk
What could go wrong, and what holds it in place.
3. The EDIE lens — people & wellbeing
How this feels for the people on the receiving end.
4. Responsible-AI checks
Tick only what is genuinely in place. Required checks must all pass before a scenario can be marked ready.
ALFI — strategy & risk

Say which part of your workload this removes, and what you will do with the time it gives back.

Record the UK GDPR basis, the accountable owner, and any DPIA or EU AI Act classification.

List each data field and delete anything you cannot justify against the purpose.

Name the three most likely failures and the three most harmful ones.

Test against work you have already marked or a plan you already trust. Set the pass mark before you look.

Name the metrics, the alert thresholds and the date of the first post-launch review.

Describe the fallback process, data retention and who authorises a shutdown.

EDIE — people & wellbeing

Check it works for pupils using a screen reader, a keyboard only, larger text and EAL support.

Ask pupils, families and support staff — including SEND leads — and record what you changed.

Think about confidence and tone. Feedback that is technically correct can still knock a pupil.

An alternative route must exist and must not be slower or poorer quality.

Write what you would say to a class and to a parent, in language they will actually use.

Shared — both cores

Say exactly where the human decision sits, who that person is, and how they can override the model.

Check results by disability, SEND status, ethnicity, sex, age and language wherever the data allows.

Say who a pupil or family goes to, and how quickly a person will reply.

5. Decision and review
The decision always belongs to the named owner. The AI review is advisory and is not legal, clinical or regulatory advice.
Add at least a couple of sentences of purpose to unlock the review.
Academy assessment and certificate
Submit a scenario that clears 70% readiness with every required check in place. Pass it and finish the course Responsible AI Scenario Design to have your CPD certificate issued and stored on your profile.
0% / 70%

Before you submit

  • Complete: Scenario title, Purpose, People affected, AI capability, Risks (ALFI), Controls (ALFI), Human impact (EDIE), Safeguards (EDIE), Accountable owner.
  • Required check outstanding: The purpose is written down and genuinely needs AI.
  • Required check outstanding: There is a lawful basis and a named owner.
  • Required check outstanding: Only the data the purpose needs is used.
  • Required check outstanding: Risks are logged with owners and severity.
  • Required check outstanding: A person decides anything that affects someone materially.
  • Required check outstanding: Outcomes are compared across affected groups.
  • Required check outstanding: The experience meets WCAG 2.2 AA and works with assistive technology.
  • Required check outstanding: The people affected have been asked, not assumed.
  • Required check outstanding: Wellbeing and dignity impacts are assessed.
  • Required check outstanding: People are told, in plain UK English, that AI is involved.
  • Required check outstanding: There is a fast route to a human when something goes wrong.
  • Required check outstanding: Live monitoring and a review date are in place.

New to this? The Academy course Responsible AI Scenario Design walks through each lens with worked examples, including SEND and NEET use cases.

Try the demo