RAI — GARP Risk and AI Certificate Study Plan
A practical study schedule for the GARP Risk and AI Certificate (RAI), with 7-day, 14-day, 30-day, and 60/90-day preparation paths.
Study Plan orientation
This Study Plan is for candidates preparing for the GARP Risk and AI Certificate (RAI), exam code RAI, from GARP. It is designed for finance, risk, audit, compliance, model validation, data, and technology professionals who need a realistic preparation schedule rather than a generic study routine.
Use this plan with the current GARP materials and learning objectives. Do not rely on a fixed topic weight unless it is stated in the official exam materials you are using. Your goal is to turn the curriculum into repeated decision practice: identifying AI risks, applying governance principles, interpreting model behavior, recognizing control gaps, and choosing appropriate risk management actions.
Which plan should you use?
| Your situation | Use this path | Minimum daily commitment | Best use of time |
|---|---|---|---|
| You have already studied most of the material and have one week left | 7-day final review | 2-4 hours | Timed practice, weak areas, error-log review |
| You know the basics but need structure quickly | 14-day focused plan | 1.5-3 hours | High-yield review plus targeted drills |
| You are starting with some finance/risk background | 30-day balanced plan | 60-120 minutes | Full coverage, spaced review, mock exams |
| You are new to AI risk or want a low-stress schedule | 60/90-day full path | 30-75 minutes | Concept building, repeated application, retention |
| Your score is inconsistent across topics | Any path, but add an error-log block daily | 20-30 extra minutes | Diagnose why answers are wrong, not just what is wrong |
Core study lanes for RAI preparation
Organize your preparation into practical study lanes. Map each lane to the current GARP RAI materials and learning objectives.
| Study lane | What to master | Practice focus |
|---|---|---|
| AI and machine learning foundations | Core terminology, model types, training concepts, model lifecycle, limitations | Distinguish similar terms and choose appropriate model-risk responses |
| Data risk and data governance | Data quality, bias, representativeness, privacy, lineage, data controls | Identify data defects and downstream risk consequences |
| Model risk management | Development, validation, approval, monitoring, documentation, change control | Apply model governance to AI and machine learning use cases |
| AI risk and control frameworks | Explainability, robustness, fairness, security, operational resilience, human oversight | Match risks to controls and escalation actions |
| Financial services applications | Credit, market, operational, fraud, AML, insurance, investment, and risk analytics use cases | Evaluate suitability, limitations, and governance needs |
| Ethics, regulation, and accountability | Responsible AI, transparency, compliance expectations, third-party risk, accountability | Interpret scenarios involving stakeholders, disclosures, and control failures |
| Quantitative and technical interpretation | Metrics, backtesting logic, validation evidence, performance drift, error trade-offs | Explain what a metric or monitoring result implies for risk decisions |
Daily practice rhythm
Use this rhythm on most study days. Short, repeated sessions are more effective than reading large sections without retrieval practice.
| Block | Time | What to do |
|---|---|---|
| Warm-up recall | 5-10 min | Write 5-10 facts, terms, risks, or controls from memory before looking at notes |
| New or review topic | 30-60 min | Study one narrow topic from the GARP materials |
| Topic drill | 20-40 min | Answer focused questions on the topic you just reviewed |
| Explanation review | 15-25 min | Read explanations for correct and incorrect answers |
| Error-log update | 10-15 min | Record missed questions by cause, not only by topic |
| Final recall | 5 min | Summarize the day in 3-5 bullet points without notes |
If you only have 45 minutes:
- Review one narrow concept for 15 minutes.
- Complete 10-15 practice questions for 20 minutes.
- Spend 10 minutes updating your missed-question log.
Missed-question review method
Do not only mark an answer as “wrong.” Identify the reason the answer was attractive and the rule that would prevent the mistake next time.
| Error type | What it means | Fix |
|---|---|---|
| Term confusion | You mixed up similar AI, risk, or governance terms | Create a two-column contrast card |
| Scenario misread | You missed the role, risk event, or control objective | Underline the decision point before answering |
| Overgeneralization | You chose a control that sounds good but does not fit the scenario | Ask: “What specific risk is being reduced?” |
| Governance gap | You knew the model concept but missed approval, monitoring, documentation, or escalation | Add lifecycle checkpoints to your notes |
| Data-risk miss | You focused on model output and ignored input data quality or bias | Trace the issue from data source to decision impact |
| Technical metric miss | You recognized the metric but not its implication | Write a plain-English interpretation of each metric |
| Compliance/ethics miss | You selected an efficient answer instead of an accountable or transparent one | Identify stakeholders, disclosures, and oversight duties |
Error-log template
Use a simple format:
| Field | Example entry |
|---|---|
| Date | June 18 |
| Topic | Data bias / representativeness |
| Question type | Scenario judgment |
| Why I missed it | Focused on model accuracy and ignored sampling bias |
| Rule to remember | Strong performance does not eliminate data governance risk |
| Retest date | In 2 days |
Review the error log every 2-3 days. In the final week, it should become your main study document.
7-day final review plan
Use this if you have one week left and have already worked through most of the GARP Risk and AI Certificate (RAI) content. If you are starting from zero, use this as a triage plan: focus on broad comprehension, scenario judgment, and high-frequency risk concepts.
| Day | Main goal | Study actions | Practice target |
|---|---|---|---|
| 1 | Diagnose readiness | Take a timed mixed quiz or short mock. Review every missed question. Sort misses by topic and error type. | 40-80 mixed questions or one timed set |
| 2 | AI/ML foundations and data risk | Review model types, data quality, bias, representativeness, privacy, lineage, and data controls. | Topic drills on terminology and data-risk scenarios |
| 3 | Model risk governance | Review development, validation, approval, monitoring, documentation, change control, and model inventory concepts. | Scenario questions on governance failures |
| 4 | Explainability, fairness, robustness, and monitoring | Focus on interpreting model behavior, drift, stability, fairness concerns, and ongoing performance monitoring. | Mixed technical interpretation questions |
| 5 | Financial services use cases | Review credit, fraud, AML, market risk, operational risk, insurance, and investment-related AI applications as relevant to your materials. | Application questions; identify risk, control, and owner |
| 6 | Timed mock and deep review | Take a timed mock or the longest realistic practice set available. Review explanations carefully. | One full mock or full-length timed simulation |
| 7 | Final consolidation | Review error log, condensed notes, formulas/metrics, governance lifecycle, and key definitions. Stop heavy new learning. | Light mixed drill only; no exhausting mock |
Final 48-hour rule
In the last two days:
- Do not add large new content areas unless they are clearly essential.
- Do not chase every obscure detail.
- Rework missed questions from your error log.
- Review definitions that change the answer in scenario questions.
- Practice reading question stems slowly and identifying the decision being tested.
- Sleep, logistics, and focus matter more than another late-night cram session.
14-day focused plan
Use this if you have two weeks and need a concentrated but realistic schedule. The plan assumes you can study most days for 1.5-3 hours.
| Day | Focus | Study tasks | Practice tasks |
|---|---|---|---|
| 1 | Baseline diagnostic | Skim the current RAI learning objectives. Take a diagnostic set. Build your topic tracker. | 40-60 mixed questions |
| 2 | AI/ML foundations | Review AI, machine learning, model types, training concepts, limitations, and terminology. | Focused terminology and concept drill |
| 3 | Data risk | Study data quality, bias, representativeness, lineage, privacy, and data controls. | Data-risk scenario questions |
| 4 | Model development lifecycle | Review development, testing, documentation, implementation, and change management. | Lifecycle sequencing questions |
| 5 | Model validation | Study validation purpose, independence, evidence, benchmarking, limitations, and challenge. | Validation and control-gap scenarios |
| 6 | Monitoring and performance | Review drift, degradation, thresholds, ongoing monitoring, and escalation. | Metric interpretation and monitoring questions |
| 7 | Checkpoint review | Retest Days 2-6. Update error log. Create one-page summaries. | Timed mixed set |
| 8 | Explainability and transparency | Review interpretability, explainability tools, stakeholder communication, and documentation. | Scenario questions on explainability needs |
| 9 | Fairness, ethics, and accountability | Study fairness, bias mitigation, responsible AI, human oversight, and accountability. | Ethics and governance scenario drill |
| 10 | Security, operational, and third-party risk | Review cyber, operational resilience, vendor models, outsourcing, and control ownership. | Third-party and operational risk questions |
| 11 | Financial services applications | Study AI use in credit, fraud, AML, trading, risk analytics, insurance, and customer decisioning as applicable. | Application-based mixed drill |
| 12 | Compliance and governance integration | Connect policy, committees, approvals, documentation, auditability, and escalation. | Governance case questions |
| 13 | Full timed practice | Take a timed mock or extended mixed set. Review deeply. | Full mock or longest available timed set |
| 14 | Final review | Review error log, weak topics, definitions, and decision rules. Avoid heavy new material. | Light retest of previously missed questions |
30-day balanced plan
Use this if you have about one month. This is the best default plan for many working candidates because it allows time for spaced repetition and at least two timed checkpoints.
30-day structure
| Phase | Days | Goal |
|---|---|---|
| Foundation | 1-7 | Build core AI, data, and risk vocabulary |
| Governance and controls | 8-15 | Understand model lifecycle, validation, monitoring, and control design |
| Application and integration | 16-22 | Apply concepts to financial services scenarios |
| Exam conditioning | 23-30 | Timed practice, error-log review, final consolidation |
Days 1-7: Foundation
| Day | Focus | Output |
|---|---|---|
| 1 | Orientation and diagnostic | Topic tracker, baseline score, weak-area list |
| 2 | AI and machine learning terminology | Definitions sheet for core terms |
| 3 | Model types and model lifecycle basics | Lifecycle map from data to monitoring |
| 4 | Data quality and data governance | Data-risk checklist |
| 5 | Bias, fairness, and representativeness | Contrast notes: bias vs fairness vs accuracy |
| 6 | Privacy, security, and data controls | Control examples linked to risks |
| 7 | Weekly review | Timed mixed set and error-log cleanup |
Days 8-15: Governance and controls
| Day | Focus | Output |
|---|---|---|
| 8 | Model risk management principles | Governance lifecycle notes |
| 9 | Model development and documentation | Documentation checklist |
| 10 | Validation and independent challenge | Validation evidence map |
| 11 | Explainability and interpretability | When explainability matters most |
| 12 | Monitoring, drift, and escalation | Monitoring trigger examples |
| 13 | Change management and implementation risk | Change-control flow |
| 14 | Third-party and vendor AI risk | Vendor risk checklist |
| 15 | Checkpoint practice | Timed mixed set; update weak-area plan |
Days 16-22: Application and integration
| Day | Focus | Output |
|---|---|---|
| 16 | Credit and underwriting applications | Risks, controls, and fairness issues |
| 17 | Fraud, AML, and surveillance applications | False positives, false negatives, monitoring |
| 18 | Market risk, trading, and investment applications | Model limitations and oversight |
| 19 | Operational risk and resilience | Control failures and escalation |
| 20 | Ethics, accountability, and transparency | Stakeholder and accountability map |
| 21 | Regulation and compliance concepts | Terminology and decision-rule review |
| 22 | Integrated scenario practice | Mixed application set and error analysis |
Days 23-30: Exam conditioning
| Day | Focus | Output |
|---|---|---|
| 23 | Full timed mock or long timed set | Baseline under time pressure |
| 24 | Mock review | Categorized error log and retest list |
| 25 | Weak area 1 | Focused review and drill |
| 26 | Weak area 2 | Focused review and drill |
| 27 | Weak area 3 | Focused review and drill |
| 28 | Second timed mock or long mixed set | Readiness check |
| 29 | Final consolidation | One-page summaries, definitions, governance lifecycle |
| 30 | Light review | Previously missed questions only; no major new topics |
60/90-day full preparation path
Use this if you are starting early, are new to AI risk, or want stronger retention. The 60-day path is a steady preparation route. The 90-day path adds more spacing, deeper notes, and more retesting.
60-day path
| Weeks | Focus | What to complete |
|---|---|---|
| Weeks 1-2 | AI, machine learning, and data foundations | Read core materials, define terms, complete topic drills |
| Weeks 3-4 | Model risk governance and validation | Study lifecycle, validation, monitoring, documentation, escalation |
| Weeks 5-6 | Responsible AI, explainability, fairness, privacy, and security | Build control maps and scenario decision rules |
| Week 7 | Financial services applications | Practice credit, fraud, AML, market, operational, and investment scenarios as applicable |
| Week 8 | Timed practice and final review | Complete mocks, review errors, consolidate notes |
90-day path
| Weeks | Focus | What to complete |
|---|---|---|
| Weeks 1-2 | Orientation and fundamentals | Read slowly, build vocabulary, take short untimed quizzes |
| Weeks 3-4 | Data risk and data governance | Focus on bias, quality, lineage, privacy, and control design |
| Weeks 5-6 | Model development and validation | Study lifecycle, validation evidence, documentation, and independent review |
| Weeks 7-8 | Monitoring, explainability, fairness, and robustness | Practice interpretation and scenario judgment |
| Weeks 9-10 | Financial services use cases | Link AI applications to risk, governance, and compliance issues |
| Week 11 | Integrated review | Mixed quizzes, weak-topic refresh, error-log retesting |
| Week 12 | Mock exams and final readiness | Timed mocks, final review, no major new content late in the week |
Weekly rhythm for 60/90-day preparation
| Day type | Activity |
|---|---|
| Study day 1 | Read or watch assigned material; make concise notes |
| Study day 2 | Practice topic questions; review explanations |
| Study day 3 | Revisit weak concepts; make comparison cards |
| Study day 4 | Apply concepts to scenarios; update error log |
| Study day 5 | Mixed quiz under light time pressure |
| Weekend or flex day | Longer review, catch-up, or timed set |
| Rest day | No heavy study; optional 10-minute flash review |
How to allocate study time by topic
Use your diagnostic results to adjust the schedule. As a default, balance conceptual learning with scenario application.
| Topic area | Suggested share of study time | Why it matters |
|---|---|---|
| AI/ML and data foundations | 20-25% | Many scenario questions depend on precise terminology and data-risk logic |
| Model risk governance and validation | 25-30% | RAI preparation should emphasize lifecycle controls, oversight, documentation, and challenge |
| Responsible AI, ethics, fairness, and explainability | 20-25% | These topics often drive judgment-based answer choices |
| Financial services applications | 15-20% | Candidates must apply AI risk concepts in finance and risk settings |
| Timed practice and error review | 15-25% | This turns knowledge into exam-ready decision making |
Adjust upward for any area where your practice accuracy is weak or your explanations are vague.
When to use timed mock exams
Timed mocks are most useful after you have enough coverage to learn from mistakes. Do not spend all your preparation time taking mocks before you can interpret the explanations.
| Preparation window | When to take timed practice | Purpose |
|---|---|---|
| 7 days | Day 1 and Day 6 if possible | Diagnose, then confirm readiness |
| 14 days | Day 7 and Day 13 | Midpoint checkpoint and final simulation |
| 30 days | Days 15, 23, and 28 if possible | Track progress and build time discipline |
| 60 days | End of Weeks 4, 6, and 8 | Move from topic knowledge to integrated judgment |
| 90 days | End of Weeks 6, 9, and 12 | Retention check and final conditioning |
Timed mock review checklist
After each mock or long timed set:
- Rework every missed question before reading the explanation.
- Mark questions you guessed correctly.
- Identify whether the miss was content, reading, timing, or judgment.
- Write one rule or trigger for each repeated error.
- Retest the same topic within 48-72 hours.
- Compare performance by topic, not just total score.
- Do not take another full mock until you have reviewed the last one.
Topic drill strategy
Use topic drills to convert reading into recall. For the GARP Risk and AI Certificate (RAI), drills should emphasize applied risk judgment rather than memorizing isolated definitions.
| Drill type | How to use it |
|---|---|
| Definition drill | Test AI, data, model risk, governance, and ethics vocabulary |
| Contrast drill | Compare similar concepts such as accuracy vs fairness, validation vs monitoring, transparency vs explainability |
| Scenario drill | Identify the primary risk, affected stakeholder, control gap, and best next action |
| Lifecycle drill | Place events in order: development, validation, approval, deployment, monitoring, change control |
| Control-matching drill | Match risk events to controls such as documentation, independent review, monitoring thresholds, escalation, or human oversight |
| Explanation drill | Explain why the correct option is better than the second-best option |
Formula, metric, and interpretation practice
RAI preparation is not only calculation practice, but candidates should be comfortable interpreting model performance, monitoring evidence, and risk trade-offs where these appear in the materials.
Create a small metric sheet with:
- What the metric measures.
- When the metric can be misleading.
- What risk decision it supports.
- What additional evidence would be needed.
- How the metric relates to fairness, drift, validation, or monitoring.
Example interpretation prompts:
| Prompt | What your answer should include |
|---|---|
| A model performs well in testing but poorly after deployment | Possible drift, data shift, monitoring failure, or implementation issue |
| A model is accurate overall but produces worse outcomes for a subgroup | Fairness, bias, representativeness, and governance review |
| A vendor model is high-performing but poorly documented | Third-party risk, explainability limits, validation constraints, and approval concerns |
| A model is retrained frequently without clear approvals | Change management, documentation, validation, and oversight issues |
Scenario-answering framework
Use the same framework for scenario questions until it becomes automatic.
- Identify the AI or model activity.
- Identify the stakeholder or business decision affected.
- Identify the primary risk: data, model, operational, compliance, ethical, security, or third-party.
- Identify the lifecycle stage: development, validation, approval, deployment, monitoring, or change.
- Choose the response that best reduces the stated risk.
- Avoid answers that are broadly positive but do not address the specific issue.
Quick decision table
| If the scenario emphasizes… | Look for an answer involving… |
|---|---|
| Poor input quality or unrepresentative samples | Data governance, remediation, lineage, bias review |
| Model cannot be explained to stakeholders | Explainability, documentation, transparency, appropriate use limits |
| Model performance changes over time | Monitoring, drift detection, thresholds, escalation |
| New model version is deployed | Change control, validation, approval, documentation |
| Vendor provides limited model details | Third-party risk management, due diligence, contractual controls |
| A subgroup is harmed or treated differently | Fairness analysis, bias mitigation, governance review |
| Business users override controls | Accountability, policy enforcement, training, escalation |
| High-impact decisioning | Stronger governance, human oversight, documentation, monitoring |
Final-week rules
During the final week, your priority is reliability. Reduce careless errors and consolidate decision rules.
| Do | Avoid |
|---|---|
| Review your error log daily | Reading large new sections passively |
| Retest missed questions | Taking mocks without reviewing them |
| Practice under time constraints | Changing your entire strategy at the last minute |
| Review definitions and contrasts | Memorizing isolated terms without context |
| Use short, focused sessions | Studying until exhausted |
| Confirm exam logistics | Leaving identification, timing, and setup to the last moment |
Exam-readiness checks
You are likely ready when most of these are true:
- You can explain key AI, machine learning, data, and model risk terms in plain language.
- You can identify the main risk in a scenario before looking at the answer choices.
- You can distinguish data risk, model risk, operational risk, compliance risk, ethical risk, and third-party risk.
- You can describe the AI/model lifecycle and where validation, approval, monitoring, and change control fit.
- Your missed questions are no longer concentrated in one or two major areas.
- You understand why your wrong answers are wrong.
- Timed practice feels controlled rather than rushed.
- Your final review notes are short enough to revisit in one sitting.
Practical next step
Choose the path that matches your available time, take a diagnostic set, and build your first error log before doing more passive reading. For RAI preparation, the highest-value routine is simple: study one topic, answer focused practice questions, review explanations carefully, and retest missed concepts under time pressure.