AAIR study plan: pass in 6 weeks

Six weeks at 6–8 hours a week, assuming you hold a qualifying designation and already work in risk.

The allocation follows the weighting, which is not where most candidates’ anxiety points. The lifecycle domain sounds the most technical and is the smallest at 21%. Programme management is 42% and governance 37% — nearly four fifths of the exam is risk discipline applied to a new asset.

Weeks 1 and 2: AI risk governance and framework integration (37%)

Week 1 — models, processes, accountability:

• AI models, frameworks, strategies and use cases. Working literacy: training versus inference, why behaviour comes from data, what a use case looks like and how its risk profile varies.
• Organisational processes and alignment — how AI adoption connects to existing governance rather than running alongside it.
• Ownership, oversight and accountability. A named accountable person, oversight with authority to intervene, and evidence it operates.

Week 2 — policy, law, ethics:

• Policies, procedures and organisational training. What an AI policy set contains, and who needs training on what.
• Regulatory compliance and legal considerations. Not clause recitation — the consistent demands: inventory, impact assessment, human oversight, documentation, transparency, records.
• Trustworthiness, ethical and societal implications, including ESG. Distinctive to AAIR; treat these as things you evidence, not values you endorse.

Week 3: AI life cycle risk management (21%)

One week for the smallest domain. Coverage, not depth.

• Design, development or procurement, and documentation — including the build-versus-buy risk trade-off.
• Model training, testing and validation — what each demonstrates and what it does not.
• Implementation, maintenance and decommissioning. Do not skip decommissioning: retiring a model raises questions about the model artefact, the training data and past decisions.
• Data and asset management — provenance, integrity, classification, retention.

Weeks 4 and 5: AI risk program management (42%)

The largest domain, two weeks.

Week 4 — identify, treat, control:

• Risk scenario identification and assessment, including threats, vulnerabilities and attacks. Learn the AI threat catalogue and which lifecycle stage each targets.
• Risk treatment strategies — mitigate, transfer, accept, avoid. Remember that documented acceptance within threshold is legitimate and frequently correct.
• Controls management: evaluation, selection and validation. Does the control address the risk, and how do you know it works?

Week 5 — measure, source, respond:

• Risk metrics, monitoring and reporting. AI thresholds are measurements — accuracy, fairness disparity, groundedness. Drift makes monitoring mandatory rather than optional.
• Supply chain risk management. Model providers, third-party datasets, APIs. Provenance, integrity, licensing, transparency, change notification and continuity.
• Incident response, business impact analysis, continuity and disaster recovery for AI. What an AI incident is, and why containment differs when behaviour lives in weights.

Week 6: practice and repair

• Full practice exam under real conditions.
• Sort mistakes by domain and by question type — a BEST question answered with a defensible-but-not-best option is a style problem, not a knowledge gap.
• Rebuild the weakest area, then sit a second practice exam.

Where the hours go

The habit that earns marks

For every risk discipline you already know, ask: what changes when the asset is a model?

Risk identification → a new threat catalogue with lifecycle stages. Assessment → probabilistic, so thresholds are measurements. Monitoring → mandatory, because drift changes risk with no trigger. Supply chain → providers who can change behaviour silently. Continuity → you cannot patch a model; containment means restricting capability. Decommissioning → the model and its past decisions outlive the service.

Six questions, six answers. Between them they cover most of the 79% that is not the lifecycle domain.

If time runs short, cut week 3. You can reason about the lifecycle from first principles; you cannot reason your way to the AI-specific content inside the risk disciplines.