FREE COURSE PROJECT · AI Risk
Management of AI Risk
HarbourHR wants an AI assistant to answer staff benefit questions from an approved knowledge base. It can draft answers but the pilot has no authority to change employee records.
Plan 9–14 hours across several sessions. Work at your own pace. Use the full Udemy course for the instructor’s complete demonstrations.
Start here
- Download and extract the pack. Open walkthrough.html in your browser to work offline.
- Read evidence-pack.html. Use only synthetic or authorized lab data.
- Record your answers in project-workbook.csv or the individual stage worksheets. Worked files include suggested decisions and limitations.
- For each task, explain what the evidence supports, what remains unknown and who should act.
- Complete the portfolio review, then continue to the full course for the complete learning sequence and its certificate of completion.
Open evidence pack · Blank project workbook · Worked examples · Completion checklist
Terms used in this project
- evidence
- An observation or record that supports a specific claim. Its source, date, scope and limitations determine what you can infer.
- risk
- The effect of uncertainty on an objective. In this project, explain a plausible event, affected service and business consequence.
- residual
- Residual risk remains after considering the controls that are actually implemented and evidenced.
- AI
- Artificial intelligence. Evaluate the actual intended use, data, model, tools, outputs and people affected.
Your practice progress
0 of 7 stages marked complete
Checklist and notes can be saved in this browser.
STAGE 1 OF 7
Frame use, impact and trustworthiness
This is how you would approach the task
- Define intended users, output, data sources and prohibited uses.
- Identify affected staff and consequences of wrong answers.
- Explain accuracy, privacy and transparency trade-offs in this use case.
What you will produce
AI use-case and impact profile.
Worked reasoning and expected decision
A confident but outdated answer could mislead a worker about benefits. The risk depends on who relies on the answer and what decision follows.
Review your result
Use the case facts to explain frame use, impact and trustworthiness. Record any missing evidence, the person who can resolve it, and what would change your conclusion.
STAGE 2 OF 7
GOVERN accountability
This is how you would approach the task
- Name service owner, data owner, risk approver and independent reviewer.
- Define pilot approval, incident escalation and change control.
- Record supplier and staff communication duties.
What you will produce
AI governance and responsibility map.
Worked reasoning and expected decision
The HR service owner owns intended use; engineering owns integration; management approves residual exposure within delegated authority.
Review your result
Use the case facts to explain govern accountability. Record any missing evidence, the person who can resolve it, and what would change your conclusion.
STAGE 3 OF 7
MAP context and failure paths
This is how you would approach the task
- Trace retrieval, model response and staff use.
- Record outdated policy retrieval and sensitive-record exposure as separate scenarios.
- Identify assumptions and populations outside the test set.
What you will produce
AI risk and dependency map.
Worked reasoning and expected decision
AI-02 can explain one plausible answer failure but does not establish the cause of all six wrong answers. Investigate before assigning a single fix.
Review your result
Use the case facts to explain map context and failure paths. Record any missing evidence, the person who can resolve it, and what would change your conclusion.
STAGE 4 OF 7
MEASURE the proposed pilot
This is how you would approach the task
- Calculate wrong answers as 6/40 and exposures as 2/40.
- Compare each with the proposed acceptance criteria.
- Record test-set coverage and sampling limitations.
What you will produce
Evaluation record and test limitations.
Worked reasoning and expected decision
Wrong answers are 15%; the target is at most 5%. Exposure is 5%; the target is zero observed exposures in the test set. Both gates fail.
Review your result
Use the case facts to explain measure the proposed pilot. Record any missing evidence, the person who can resolve it, and what would change your conclusion.
STAGE 5 OF 7
MANAGE treatment and profiles
This is how you would approach the task
- Prioritize removal of sensitive retrieval sources and outdated pages.
- Define human review and a repeat test set.
- Compare current and target outcomes and document the pilot decision.
What you will produce
Treatment plan and current/target profile.
Worked reasoning and expected decision
Defer broad launch. A later clean test supports a scoped pilot decision but cannot prove the absence of all future failures.
Review your result
Use the case facts to explain manage treatment and profiles. Record any missing evidence, the person who can resolve it, and what would change your conclusion.
STAGE 6 OF 7
Use Playbook, crosswalk and SAIF
This is how you would approach the task
- Select applicable NIST AI RMF guidance for each recorded scenario.
- Compare the AI application against Google SAIF security considerations.
- Identify a security-specific test missing from the broader impact review.
What you will produce
Framework crosswalk and security review.
Worked reasoning and expected decision
A risk framework and a technical security review answer related questions. Keep their evidence linked without treating one as proof of the other.
Review your result
Use the case facts to explain use playbook, crosswalk and saif. Record any missing evidence, the person who can resolve it, and what would change your conclusion.
STAGE 7 OF 7
Present the approval recommendation
This is how you would approach the task
- Write a one-page decision with results, owners, remaining gaps and pause trigger.
- Set retest and material-change review dates.
- Explain the intended pilot to staff.
What you will produce
AI approval brief and monitoring plan.
Worked reasoning and expected decision
Approval depends on corrected retrieval boundaries, independent retest and accountable oversight. Provide the exact decision the sponsor needs to make.
Review your result
Use the case facts to explain present the approval recommendation. Record any missing evidence, the person who can resolve it, and what would change your conclusion.
Course coverage and learning review
Course topics checked against the published curriculum returned by Udemy on 9 November 2026. Lecture titles are a coverage index; the worked project is an original exercise.
The topics below identify how this exercise connects to the course. After each section, explain how its concepts change a decision in the case. Record examples or questions that require further study in curriculum-review.csv.
PART 1
- The Importance of AI Risk Management · Curriculum reflection
- 1-Course Introduction · Curriculum reflection
- 2-Intro the Framework outline · Curriculum reflection
- 3-AI RMF Executive Summary pt1 · Curriculum reflection
- 4-AI RMF Executive Summary pt2 · Curriculum reflection
- 5-Part 1-Framing Risk pt1 · Curriculum reflection
- 6-Part 1-Framing Risk pt2 · Curriculum reflection
- 7-Part 1-Framing Risk pt3 · Curriculum reflection
- 8-Part 1-Audience · Curriculum reflection
- 9-Part 1-AI Risks and Trustworthiness pt1 · Curriculum reflection
- 10-Part 1-AI Risks and Trustworthiness pt2 · Curriculum reflection
- 11-Part 1-AI Risks and Trustworthiness pt3 · Curriculum reflection
- 12-Part 1-AI Risks and Trustworthiness pt4 · Curriculum reflection
- 13-Part 1-AI Risks and Trustworthiness pt5 · Curriculum reflection
- 14-Part 1-AI Risks and Trustworthiness pt6 · Curriculum reflection
- 15-Part 1-AI Risks and Trustworthiness pt6 · Curriculum reflection
- 16-Part 1-AI Risks and Trustworthiness pt7 · Curriculum reflection
- 17-Part 1-Effectiveness of the AI RMF · Curriculum reflection
PART 2
- 18-Intro to Part 2-Core and Profiles · Curriculum reflection
- 19-AI RMF Core-Govern · Curriculum reflection
- 20-AI RMF Core-Map · Curriculum reflection
- 21-AI RMF Core-Measure · Curriculum reflection
- 22-AI RMF Core-Manage · Curriculum reflection
- 23-AI RMF Profiles · Curriculum reflection
PART 3
- 24- The AI RMF Appendix Documents · Curriculum reflection
- 25-The AI RMF Tables Summary for each function · Curriculum reflection
- 26- The AI RMF Roadmap · Curriculum reflection
- 27- The AI RMF Playbook · Curriculum reflection
- Application of The NIST AI 100-1 Risk Management Framework (RMF) · Related stage 6
- 28-Case Scenario-Requirements and Required Actions · Curriculum reflection
- 29-Case Scenarion Report Presentation-pt1 · Curriculum reflection
- 30-Case Scenarion Report Presentation-pt2 · Curriculum reflection
- 31-NIST RMF Crosswalk Documents · Curriculum reflection
PART 4
- Intro _Conducting a Practical AI Risk Self Assessment using Google SAIF · Related stage 6
- Exploring the elements of Google SAIF · Curriculum reflection
- Using Google SAIF to Conduct a Practical AI Risk Self Assessment · Related stage 6
- End of Course · Curriculum reflection
References and related CyberJA tools
Use these sources to deepen the reasoning behind the tasks. Check the current version and applicability before applying guidance to a real organization.
Build your portfolio
- Assemble the stated deliverable from every stage, with record IDs linking claims to synthetic evidence.
- Write a one-page recommendation that names the decision, evidence, assumptions, owners and next review.
- Prepare a two-minute explanation of a conclusion you changed when the evidence was incomplete.
- Label the work as a simulated project in your CV or professional profile.
Self-assessment rubric
Score each area from 0 to 4: 0 absent, 1 assertion without support, 2 partial evidence, 3 supported and reproducible, 4 supported with limitations and alternatives. Aim for at least 3 in every area.
- Correct use of the scenario and source facts.
- Clear reasoning and treatment of uncertainty.
- Practical implementation or decision steps.
- Evidence, ownership and verification.
- Clear communication and honest portfolio claims.
This rubric is a self-review aid. Your checklist is self-recorded.
Continue with the full Udemy course
Use your completed practice work as a starting point for Richea Perry’s full course, Management of AI Risk. Learn through the instructor’s complete demonstrations and course resources.
Enroll and complete the required curriculum items on Udemy to receive the course’s certificate of completion. This free project records practice progress. The certificate is issued through completion of the full Udemy course.
Take the full course on Udemy ↗Udemy certificate requirements · Discuss your practice in CyberJA Connect