Only completion ticks are saved in this browser. No account, analytics or learner submissions.
LESSON 01
Inventory tools and data boundaries
Ask a team owner to list approved AI tools, vendors, intended users, company-controlled accounts, sensitive data types and current approval rules. Classify examples as public, internal or sensitive per your company policy. Do not upload personal records to test a tool.
Do this
Complete tool-register.csv. For each tool include its allowed inputs, forbidden inputs, reviewer and status (APPROVED/REVIEW/HOLD).
Completion check: Your list distinguishes approved tools from tools merely used informally.
LESSON 02
Evaluate one use case at a time
For a proposed workflow, record purpose, value, data source, possible wrong answer, unauthorized action risk and how a human checks the output. Determine whether a deterministic rule or manual step would be safer. Stop if you lack permission.
Do this
Fill use-case-review.md for two fictional use cases: summarizing public release notes and processing a confidential employee file. Reject or hold the second until proper authority and controls exist.
Completion check: Every candidate use case has an explicit owner, risk rating, reviewer and allow/hold decision.
LESSON 03
Practice three failure scenarios
Give the team fictional examples: an invented citation, a request to upload private employee details, and an assistant offering to email external contacts automatically. The default is to refuse unsafe input and escalate actions outside authorization.
Do this
Fill scenario-checks.md with observed failure, stop step, owner and remediation. Compare responses from two colleagues before approving a workflow.
Completion check: At least three scenario responses identify when a human must intervene.
LESSON 04
Hold a documented team review
Walk through the approved use cases and rules in a short training meeting. Record attendees, materials, questions and any open risks. Decide on an owner and next review date. This kit is a starter training aid, not an ISO or EU AI Act certification.
Do this
Complete training-record.csv, sign-off checklist and 30-day follow-up. Do not use this record as evidence of full regulatory compliance.
Completion check: A team lead can show who reviewed the process, what remains on HOLD and when it is revisited.
Reusable prompt (human review required)
Review this *fictional* proposed AI workflow for data sensitivity, inaccurate output, permission boundaries, human review, and record retention. State an ALLOW, HOLD or REJECT recommendation for human decision. Do not claim legal or regulatory certification. List open questions.
Fictional worked example
Example only: Team wishes to summarize public changelogs and also upload real personnel files. Public changelogs can enter a limited reviewed trial; personnel-file processing is HOLD pending privacy and company authorization. These are fictional training decisions.
Take the working toolkit
Includes a complete field guide PDF, four editable templates, a prompt, fictional example and completion checklist. All files work offline after download.
