Use AI without outsourcing judgment
Create one reusable, privacy-aware AI workflow that improves real work and includes verification.
Choose tasks where AI is useful and risk is manageable.
Write instructions with context, constraints, and a quality bar.
Verify claims and protect sensitive information.
Make the example fit your situation.
Choose a practice context. This choice does not ask for or save your age.
Use synthetic or permitted non-confidential data to improve a repeatable work task. A person verifies sources and owns the final decision.
Develop AI literacy as an accountable work process, not prompt memorization. The project makes task choice, privacy, verification, and human responsibility inspectable.
Reuse your lesson work in the project. Optional videos and extended activities add time; these are planning estimates.
30 reading + 141 practice/checks + 55 additional project work + 25 weekly review
Do the work in this order
- 01
Choose a low-risk, repeatable task with a clear human owner.
- 02
Define allowed inputs, protected data, constraints, and the quality standard before prompting.
- 03
Run the workflow on a fresh example and log every source or calculation that requires verification.
- 04
Insert one deliberately flawed output and confirm the stated control catches it.
The learner catches the seeded failure, verifies material claims outside the model, and can state who owns the final decision.
Complete the lesson sequence
Pass each lesson check and your lesson progress saves automatically. Return here to see your progress and work on the weekly project.
- 1aiFind the right work for AIStart
Why this lesson is here: Choose one repeatable task for your AI work recipe whose inputs are allowed, errors are manageable, and results you can independently check.
- 2aiWrite strong instructionsStart
Why this lesson is here: Write the reusable instructions for that recipe, including its audience, source facts, output format, constraints, and what to do when information is missing.
- 3aiBreak complex work into stepsStart
Why this lesson is here: Break your chosen AI task into clear steps with inputs, outputs, and human checkpoints so you can locate and correct a failure.
- 4aiGround and verify outputsStart
Why this lesson is here: Add source and calculation checks to the same AI recipe, then record a correction that shows how you verified an output.
- 5aiProtect data and rightsStart
Why this lesson is here: Name allowed inputs, excluded information, and permission checks for your AI recipe before using any data.
- 6aiBuild a workflow libraryStart
Why this lesson is here: Package your one AI recipe so it can be reused, with an example, test notes, failure conditions, and a named person responsible for the final decision.
Try the workflow on a new example and deliberately introduce one bad output. The stated checks should catch it.
Verified AI work recipe
A reusable workflow with inputs, instructions, checkpoints, sources, and a human approval step.
Describe the task, acceptable inputs, protected data, prompt or instructions, decomposition steps, verification method, failure conditions, and final human decision. Log 2 tests: a fresh example and a deliberately flawed output that your checks must catch.
- Document 1 reusable AI work recipe with a task, quality bar, permitted inputs, and clear instructions or stages.
- Define privacy limits, source and calculation checks, failure conditions, and the person who owns the final decision.
- Log 2 tests: a fresh example and a deliberately flawed output that your checks must catch.
- Record results and revisions; do not claim the workflow is verified beyond what these limited tests show.
Task and quality bar are specific
Sensitive data is excluded
Important claims are independently checked
A person owns the final decision