Lessons and outlineLesson 10 of 10
AI Governance & Risk LeadLesson 10
Governance Interviews & Case Practice
Common questions with strong answers, a method for cases and a portfolio that proves your skill.
Interviewers look for plain explanation, proportionate programs, framework knowledge, testable controls and honest limits. Use a method: clarify, tier, rules, risks, controls, decision, monitor and people. Show a policy, an inventory, a mock review, an impact assessment and a 90-day plan.
- Decision memo
- A short note recording a decision, its reasons and its conditions.
- Proportionate
- Matching the effort to the size of the risk.
- First 90 days
- The plan for a new role's first three months.
This final lesson prepares you for the interview. Picture the hiring manager. They have a growing number of AI projects and little structure. They want someone who can build the program: practical, calm and able to work with engineers and lawyers alike. They will test your judgment on cases. This lesson shows common questions, a method and a portfolio.
What interviewers look for
- You can explain AI risk in plain words to executives and engineers.
- You build proportionate programs: light for low risk, deep for high risk.
- You know the main frameworks and laws, and you are careful about legal advice.
- You can design controls that are testable and produce evidence.
- You balance speed and safety, and you are easy to work with.
- You are honest about uncertainty.
Common questions and strong answers
Question: you join a company with no AI governance. What do you do in the first 90 days? A strong answer starts with discovery: find out what AI is already in use, by survey and spend review. It builds a simple inventory and a draft risk tier scheme. It writes a short policy with allowed tools and data rules. It offers fast approval for low-risk tools. It picks two high-risk uses for a deeper review. It reports early findings to leaders.
Question: a team wants to use AI to screen job applicants. What is your view? A strong answer treats it as high risk. It asks about the purpose, data, fairness testing and human oversight. It points to laws on hiring and automated decisions in the relevant region. It requires an impact assessment, bias testing across groups, a way for candidates to reach a person and a decision by a senior owner. It does not say never. It says what must be true first.
Question: engineers say your reviews slow them down. What do you do? A strong answer takes it seriously. It looks at review times by tier, removes steps for low-risk uses, offers templates and pre-approved patterns, and puts a named contact on call. It shares what the reviews have caught, and asks engineers to help design the process.
Question: how would you know that human oversight really works? A strong answer tests it. It samples approvals, plants known errors, checks review times and volumes and talks to reviewers. It looks for blind trust and rubber stamps, and it fixes volume, training and tools.
A method for governance cases
1. CLARIFY What is the system, who uses it, who is affected, what data, what decision? 2. TIER How risky is it? Which questions set the tier? 3. RULES Which laws, standards, contracts and policies may apply? What needs legal advice? 4. RISKS Privacy, fairness, safety, security, IP, vendor, reputation. Rate them. 5. CONTROLS Preventive, detective, corrective. Who owns each? How is it tested? 6. DECISION Go, go with conditions, or stop. Who decides? 7. MONITOR What do you watch after launch? When is the next review? 8. PEOPLE Who needs training? How do you keep the process easy to use?
Show your work
- Write a sample AI policy of two pages, a risk tier scheme and an inventory template.
- Do a mock review of a made-up system: a risk assessment, controls and a decision memo.
- Write a short impact assessment for a high-risk use, such as a hiring tool.
- Draft a vendor questionnaire and a contract checklist.
- Prepare a one-page 90-day plan for a company with no program.
Questions to ask them
- What AI is in use today, and who owns it?
- Where does governance report, and who can say stop?
- What has gone wrong so far, and what changed?
- How do you balance speed with safety on new launches?
- ✓Interviewers look for plain explanation, proportionate programs, framework knowledge, testable controls and honest limits.
- ✓Be ready for the first-90-days question, a hiring-tool review, pushback from engineers and testing human oversight.
- ✓Use a method: clarify, tier, rules, risks, controls, decision, monitor and people.
- ✓Show a policy, an inventory, a mock review, an impact assessment, a vendor checklist and a 90-day plan.
Answer a case aloud
- Pick this case: a bank wants an AI assistant to help staff decide on small business loans.
- Answer in ten minutes using the eight steps. Record yourself.
- Listen back. Mark where you skipped the tier, the law check or monitoring.
- Write a one-page decision memo with your conditions for go.
Q1A team wants to use AI to screen job applicants. What is your first reaction?show answer
Treat it as high risk, not an automatic no. Ask about purpose, data, bias testing, human oversight and the law, and require an impact assessment and a senior decision.
Q2Engineers say reviews slow them down. What do you do?show answer
Look at review times by tier, remove steps for low-risk uses, add templates and pre-approved patterns, and involve engineers in designing the process.
Q3How do you show that human oversight really works?show answer
Sample approvals, plant known errors, check time and volume and talk to reviewers. Fix blind trust with lower volume, training and better tools.