AI Literacy and Responsible Practice

Compare predictive, generative and robotic systems using everyday examples.

Evidence: Six examples categorised with realistic capabilities and limitations.

Trace collection, preparation, training, deployment and monitoring.

Evidence: A lifecycle map identifying a responsible owner at each stage.

Write a bounded instruction with context and success criteria.

Evidence: Two prompt revisions justified by observable output differences.

Verify a generated answer against a supplied source pack.

Evidence: All factual claims are traced or marked unsupported.

Inspect a dataset for missing groups and unequal errors.

Evidence: A subgroup comparison and a proposed mitigation with tradeoffs.

Identify sensitive fields and choose synthetic or minimised substitutes.

Evidence: A redacted example that retains task usefulness without personal details.

Record source, licence status and permitted use for supplied assets.

Evidence: A provenance register; unresolved permissions are flagged for review.

Design approval points for higher-impact AI actions.

Evidence: A workflow that routes uncertain or consequential outputs to a person.

Recognise prompt injection and unsafe tool requests in examples.

Evidence: A classification exercise with explanations for all unsafe actions.

Define quality, time and cost measures for a proposed AI pilot.

Evidence: A baseline and measurable success criteria rather than vendor claims.

Explain an AI-assisted result to a nontechnical stakeholder.

Evidence: A short briefing that states evidence, uncertainty and accountability.

Sequence a small pilot, review and controlled rollout.

Evidence: A plan with an owner, stop conditions and a review date.