AI assurance practice library32
PlanMAN-01
AI assurance charter
Define internal audit authority, scope, access, independence, assurance responsibilities, advisory boundaries, reporting lines, and escalation rights for AI.
FieldworkMAN-02
Independence and objectivity safeguards
Identify prior advisory work, management responsibilities, specialist relationships, conflicts, and safeguards before providing assurance.
ReportMAN-03
Board and audit committee oversight
Set reporting expectations for AI risk coverage, significant findings, accepted risk, overdue actions, capability, and assurance limitations.
Follow-upMAN-04
Three-lines assurance coordination
Coordinate business monitoring, risk and compliance testing, model validation, security assurance, external review, and internal audit reliance.
PlanUNI-01
AI system universe linkage
Link auditable entities to use cases, systems, models, agents, data, vendors, processes, products, jurisdictions, owners, and customer impacts.
FieldworkUNI-02
Assurance risk assessment
Assess consequence, complexity, autonomy, scale, change, control maturity, incidents, complaints, vendor dependence, and evidence quality.
ReportUNI-03
Coverage and reliance map
Show which material risks and systems are covered, by whom, with what scope, frequency, independence, quality, and remaining assurance gaps.
Follow-upUNI-04
Dynamic audit plan
Prioritize engagements and continuous assurance using risk, change, regulatory interest, strategic importance, incidents, control signals, and capacity.
PlanPLN-01
Engagement objective and criteria
State the assurance question, scope, standards, policies, laws, risk appetite, control criteria, period, and expected form of conclusion.
FieldworkPLN-02
Use-case and system boundary
Trace the full socio-technical system including data, models, prompts, tools, integrations, people, vendors, decisions, monitoring, and recourse.
ReportPLN-03
Evidence and sampling plan
Define populations, samples, periods, sources, completeness checks, reproducibility, exceptions, data handling, and retention before fieldwork.
Follow-upPLN-04
Specialist and access plan
Secure appropriate insurance, actuarial, data, model, security, legal, accessibility, fairness, and technical expertise plus timely system access.
PlanGOV-01
Policy and accountability test
Test whether approved policy, ownership, decision rights, training, escalation, oversight, and risk acceptance operate as designed.
FieldworkGOV-02
Inventory and risk-tiering test
Reconcile production reality to the inventory and verify classification, dependencies, ownership, review cadence, and control proportionality.
ReportGOV-03
Stage-gate and approval test
Inspect whether design, validation, release, change, and retirement decisions used complete evidence, qualified challenge, and proper authority.
Follow-upGOV-04
Change and retirement control test
Test material-change detection, revalidation, approvals, rollback, migration, data disposition, access removal, records, and closure.
PlanTEC-01
Data rights quality and lineage test
Test provenance, permission, quality, representativeness, currency, transformation, lineage, access, retention, and downstream correction.
FieldworkTEC-02
Model and agent validation review
Assess validation independence, methods, test sets, benchmarks, uncertainty, limitations, robustness, autonomy, tool use, and production relevance.
ReportTEC-03
Security resilience and tool-access test
Test identity, permissions, secrets, logging, prompt injection, data leakage, supply chain, misuse, recovery, and safe-failure controls.
Follow-upTEC-04
Explainability fairness and oversight test
Evaluate explanation fidelity, outcome disparities, human authority, override, escalation, accessibility, documentation, and consumer recourse.
PlanOPS-01
Production monitoring test
Test metric design, data completeness, thresholds, alerting, ownership, investigation, drift, quality, use, value, and corrective action.
FieldworkOPS-02
Incident complaint and appeal test
Trace incidents, complaints, corrections, reviews, appeals, root causes, notifications, remediation, recurrence, and lessons learned.
ReportOPS-03
Consumer outcome testing
Evaluate accuracy, service, coverage, pricing, claims, fairness, accessibility, explanations, and recourse across relevant populations and contexts.
Follow-upOPS-04
Vendor and service control test
Assess contracts, evidence, service levels, changes, incidents, subcontractors, concentration, audit rights, continuity, and exit readiness.
PlanFND-01
Finding classification and root cause
Describe condition, criteria, cause, effect, scope, evidence, risk, recurrence, ownership, and urgency consistently.
FieldworkFND-02
Management action and remediation
Require specific actions that address root cause with accountable owners, milestones, resources, due dates, success criteria, and interim controls.
ReportFND-03
Validation and closure
Independently verify implementation, operating evidence, effectiveness, sustainability, and residual risk before closing a finding.
Follow-upFND-04
Overdue escalation and risk acceptance
Escalate missed actions and ensure material residual risk is accepted only by authorized management and reported appropriately.
PlanRPT-01
Engagement report
Communicate scope, criteria, work performed, limitations, conclusion, significant observations, findings, management response, and follow-up.
FieldworkRPT-02
Portfolio assurance opinion
Aggregate engagement and other-assurance evidence to describe coverage, themes, systemic issues, confidence, gaps, and risk trends.
ReportRPT-03
Regulatory examination readiness
Maintain traceable evidence, issue history, validated closure, accountable responses, and a coordinated process for examination requests.
Follow-upRPT-04
Quality and methodology improvement
Review supervision, evidence, consistency, stakeholder feedback, specialist use, analytics, lessons, and conformance to improve AI assurance.