1. Choose a safe first task
- Start with a task you understand and can check.
- Use AI to assist—not to make final high-impact decisions.
- Remove private, confidential and unnecessary personal information.
- Decide who must review the result before it is used.
2. Build a CLEAR prompt
Context: The situation, task and background.
Limits: Boundaries, safety, privacy and what not to invent.
Examples or evidence: Approved source information or a model.
Audience and accessibility: Who it is for, plain language, structure and access needs.
Result: The required format, length and next action.
3. Verify the answer
- Check names, dates, numbers, links and quotations.
- Separate sourced facts from the AI’s interpretation.
- Ask for uncertainty to be stated clearly.
- Check for bias, exclusion, stereotypes and inaccessible structure.
- Use an authoritative source for important claims.
- Keep a record of what a person reviewed or changed.
Stop and get specialist help when
The task could materially affect health, safety, rights, employment, legal status, finances, benefits, education access or another high-impact outcome. AI may assist research or drafting, but an accountable qualified person must lead the decision.
My first-task record
Task: ______________________________________________
Why AI is suitable: __________________________________
Information removed or protected: ______________________
Who will review the result: _____________________________
What I checked and changed: ____________________________
Accessibility Compliance Note
This pack uses semantic headings, plain language, strong contrast, meaningful labels and a simple print layout. It can be printed, saved as a PDF or read directly with assistive technology.