> ## Documentation Index
> Fetch the complete documentation index at: https://docs.aigoframework.com/llms.txt
> Use this file to discover all available pages before exploring further.

# 10 AIGO AI Assurance Procedure v0.1

# AIGO — AI Governance Operating Framework

## AI Assurance Procedure

**Version:** 0.1\
**Status:** Draft\
**Working Name:** AIGO\
**Full Name:** AI Governance Operating Framework\
**Document Identifier:** AIGO-PROC-010

***

### 1. Purpose

This procedure defines the process for planning, conducting, documenting, reporting, and following up on assurance activities for AI systems and the AIGO governance framework.

The procedure provides a structured approach for determining whether AI governance requirements, controls, processes, and system-level safeguards are appropriately designed, implemented, and operating effectively.

***

### 2. Scope

This procedure applies to assurance activities within the organization's AIGO governance scope.

It may apply to:

* AI governance processes;
* AI systems;
* AI models;
* AI applications;
* generative AI systems;
* AI agents;
* AI controls;
* risk management processes;
* lifecycle processes;
* monitoring activities;
* third-party AI services; and
* supporting governance mechanisms.

***

### 3. Objectives

The objectives of AI assurance are to:

* provide confidence in governance effectiveness;
* evaluate control design;
* evaluate control implementation;
* evaluate control operation;
* identify weaknesses;
* identify gaps;
* verify remediation;
* support governance decisions; and
* promote continuous improvement.

***

### 4. Assurance Principles

AI assurance should be:

* objective;
* evidence-based;
* risk-based;
* proportionate;
* independent where required;
* repeatable;
* documented;
* traceable; and
* transparent about limitations.

***

### 5. Assurance and Audit

Assurance and audit are related but distinct activities.

Assurance may provide confidence over:

* governance;
* controls;
* risk management;
* AI system performance;
* compliance;
* operational processes; and
* implementation.

Internal or external audit may form part of an organization's broader assurance model.

***

### 6. Assurance Scope

The assurance scope should clearly define:

* subject;
* objectives;
* criteria;
* period;
* systems;
* controls;
* stakeholders;
* evidence; and
* limitations.

***

### 7. Assurance Triggers

Assurance may be initiated by:

* risk level;
* AI system classification;
* regulatory requirements;
* governance requirements;
* approval conditions;
* material changes;
* incidents;
* monitoring findings;
* audit findings;
* management request; or
* periodic assurance planning.

***

### 8. Assurance Planning

The organization should establish an assurance plan appropriate to its AI governance environment.

The plan should consider:

* AI system risk;
* governance maturity;
* previous findings;
* incidents;
* changes;
* control effectiveness;
* regulatory requirements; and
* assurance priorities.

***

### 9. Assurance Types

Assurance activities may include:

* governance assessment;
* control assessment;
* compliance assessment;
* technical assessment;
* model validation;
* security assessment;
* privacy assessment;
* process assessment;
* operational review;
* independent review; and
* audit.

***

### 10. Assurance Levels

The depth of assurance should be proportionate to the subject's risk and significance.

Possible levels include:

1. Basic review.
2. Structured assessment.
3. Detailed assurance review.
4. Independent assurance.
5. External assurance.

***

### 11. Assurance Criteria

Assurance should be performed against defined criteria.

Criteria may include:

* AIGO requirements;
* organizational policies;
* procedures;
* controls;
* contractual requirements;
* legal requirements;
* regulatory requirements;
* technical requirements; and
* approved system specifications.

***

### 12. Assurance Evidence

Assurance conclusions should be supported by sufficient and appropriate evidence.

Evidence may include:

* policies;
* procedures;
* risk assessments;
* control records;
* system documentation;
* configuration;
* logs;
* monitoring records;
* testing results;
* approvals;
* incident records; and
* interviews.

***

### 13. Evidence Sufficiency

Evidence should be sufficient to support the assurance conclusion.

Evidence should be evaluated for:

* completeness;
* relevance;
* reliability;
* accuracy;
* timeliness; and
* traceability.

***

### 14. Evidence Limitations

Assurance reports should identify significant evidence limitations.

Limitations may include:

* unavailable records;
* incomplete data;
* restricted system access;
* third-party dependencies;
* insufficient historical information; and
* sampling limitations.

***

### 15. Assurance Independence

Where independence is required, assurance should be performed by personnel who are sufficiently independent from the activity being assessed.

Independence requirements should be proportionate to:

* risk;
* assurance purpose;
* organizational governance;
* regulatory requirements; and
* stakeholder expectations.

***

### 16. Conflict of Interest

Assurance personnel should disclose relevant conflicts of interest.

Where a conflict may impair objectivity, alternative assurance arrangements should be considered.

***

### 17. Assurance Planning Criteria

Before beginning an assurance activity, the assurance owner should define:

* objective;
* scope;
* criteria;
* methodology;
* evidence requirements;
* responsible personnel;
* timeline; and
* reporting approach.

***

### 18. Assurance Methods

Assurance methods may include:

* document review;
* interviews;
* observation;
* sampling;
* testing;
* technical analysis;
* control walkthroughs;
* data analysis;
* system inspection; and
* independent validation.

***

### 19. Sampling

Where sampling is used, the sample should be appropriate to the assurance objective.

Sampling considerations may include:

* population;
* risk;
* materiality;
* frequency;
* system diversity; and
* historical findings.

***

### 20. Control Design Assessment

Assurance may assess whether controls are appropriately designed to address identified risks.

The assessment should consider:

* control objective;
* risk addressed;
* control mechanism;
* responsibility;
* frequency;
* evidence; and
* expected outcome.

***

### 21. Control Implementation Assessment

Assurance should determine whether required controls have been implemented.

Evidence may include:

* configuration;
* procedures;
* assigned ownership;
* system functionality;
* records; and
* operational evidence.

***

### 22. Control Operating Effectiveness

Where applicable, assurance should determine whether controls operate as intended over the relevant period.

Assessment may consider:

* execution;
* consistency;
* exceptions;
* evidence;
* failures; and
* corrective actions.

***

### 23. AI System Assurance

AI system assurance may consider:

* intended purpose;
* classification;
* risk;
* performance;
* robustness;
* security;
* privacy;
* fairness;
* safety;
* human oversight;
* monitoring; and
* controls.

***

### 24. Model Assurance

Where applicable, model assurance may assess:

* model design;
* training data;
* validation;
* performance;
* robustness;
* limitations;
* drift;
* explainability;
* documentation; and
* monitoring.

***

### 25. Generative AI Assurance

Generative AI assurance may consider:

* output quality;
* hallucination;
* harmful content;
* prompt injection;
* data leakage;
* access controls;
* system instructions;
* monitoring;
* human oversight; and
* misuse controls.

***

### 26. AI Agent Assurance

AI agent assurance may consider:

* autonomy;
* permissions;
* tools;
* actions;
* transaction authority;
* human oversight;
* logging;
* escalation;
* safeguards; and
* action reversibility.

***

### 27. Third-Party Assurance

Third-party AI services should be subject to appropriate assurance.

Assurance may consider:

* provider controls;
* contractual commitments;
* certifications;
* independent reports;
* security;
* privacy;
* service performance;
* model changes; and
* incident history.

***

### 28. Assurance Findings

Findings should be documented where evidence indicates:

* non-conformity;
* control weakness;
* governance gap;
* ineffective control;
* insufficient evidence;
* material risk; or
* improvement opportunity.

***

### 29. Finding Classification

Findings may be categorized according to significance.

A representative classification is:

1. Critical.
2. High.
3. Moderate.
4. Low.
5. Observation.
6. Improvement opportunity.

The organization should define appropriate criteria for each category.

***

### 30. Finding Criteria

Finding significance may consider:

* risk;
* impact;
* likelihood;
* control failure;
* affected population;
* regulatory significance;
* recurrence;
* duration; and
* management response.

***

### 31. Finding Root Cause

Material findings should identify root causes where appropriate.

Root causes may include:

* inadequate governance;
* unclear responsibility;
* insufficient controls;
* process failure;
* technical weakness;
* inadequate training;
* monitoring failure; or
* resource constraints.

***

### 32. Corrective Actions

Corrective actions should address identified findings.

Each action should identify:

* finding;
* action;
* owner;
* priority;
* target date;
* evidence;
* verification method; and
* closure criteria.

***

### 33. Remediation Tracking

Material findings should be tracked until closure.

Tracking should identify:

* status;
* owner;
* due date;
* evidence;
* dependencies;
* delays;
* escalation; and
* closure.

***

### 34. Finding Acceptance

Where permitted, a finding may be accepted as residual risk by an authorized risk owner.

Acceptance should not remove the finding from the assurance record.

***

### 35. Assurance Report

The assurance activity should produce a report appropriate to its scope.

The report should include:

* objective;
* scope;
* criteria;
* methodology;
* evidence;
* findings;
* limitations;
* conclusions; and
* recommendations.

***

### 36. Assurance Conclusion

The assurance conclusion should clearly communicate the level of confidence supported by the evidence.

Possible conclusions may include:

1. Effective.
2. Generally effective with improvements required.
3. Partially effective.
4. Ineffective.
5. Unable to conclude.

The organization should define the formal meaning of each conclusion.

***

### 37. Management Response

Responsible management should provide responses to material findings.

Responses should identify:

* action;
* owner;
* target date;
* priority;
* resources;
* dependencies; and
* acceptance where applicable.

***

### 38. Assurance Escalation

Findings should be escalated when:

* risk exceeds tolerance;
* remediation is overdue;
* management does not respond;
* repeated findings occur;
* critical controls fail;
* significant incidents occur; or
* regulatory exposure exists.

***

### 39. Assurance Follow-Up

Follow-up assurance should determine whether corrective actions have been effectively implemented.

Follow-up may include:

* evidence review;
* retesting;
* control testing;
* system review;
* interviews; and
* independent verification.

***

### 40. Assurance Closure

An assurance finding should be closed only when:

* required action is completed;
* sufficient evidence exists;
* effectiveness has been verified where required; and
* closure is authorized.

***

### 41. Repeated Findings

Repeated findings should be analyzed for systemic causes.

Repeated findings may indicate:

* ineffective remediation;
* inadequate ownership;
* insufficient resources;
* weak governance;
* inadequate controls; or
* inadequate monitoring.

***

### 42. Assurance Reporting

The AI governance function should report assurance results to appropriate governance bodies.

Reporting may include:

* assurance activities completed;
* findings;
* severity;
* overdue actions;
* recurring findings;
* control effectiveness;
* emerging risks; and
* overall assurance conclusions.

***

### 43. Assurance Metrics

Organizations may establish assurance metrics.

Examples include:

* assurance coverage;
* findings by severity;
* overdue findings;
* repeat findings;
* remediation cycle time;
* control effectiveness;
* assurance completion rate; and
* unresolved high-risk findings.

***

### 44. Assurance Records

Assurance records should be maintained according to applicable retention requirements.

Records may include:

* assurance plan;
* scope;
* criteria;
* evidence;
* workpapers;
* findings;
* reports;
* management responses;
* remediation evidence; and
* closure records.

***

### 45. Assurance Traceability

Assurance records should maintain traceability to relevant governance artifacts.

The organization should be able to demonstrate:

**AI System → Risk → Controls → Evidence → Assurance → Findings → Remediation → Verification**

***

### 46. Assurance and Risk Management

Assurance findings should feed into AI risk management.

Material findings should trigger risk reassessment where appropriate.

***

### 47. Assurance and Control Management

Assurance should provide information about control effectiveness.

Control weaknesses should be reflected in relevant control records and remediation plans.

***

### 48. Assurance and Incident Management

Significant incidents should be considered in assurance planning.

Incident findings may trigger targeted assurance activities.

***

### 49. Assurance and Change Management

Material changes should be considered when determining assurance requirements.

Changes may require:

* additional assurance;
* targeted testing;
* independent review; or
* post-change assurance.

***

### 50. Assurance Responsibilities

**Assurance Owner**

* define assurance scope;
* establish criteria;
* coordinate assessment;
* evaluate evidence;
* document findings; and
* issue conclusions.

**AI Governance Function**

* maintain assurance governance;
* coordinate assurance planning;
* monitor findings;
* escalate material issues; and
* report assurance status.

**AI System Owner**

* provide evidence;
* support assessment;
* respond to findings;
* implement corrective actions; and
* provide remediation evidence.

**Specialist Functions**

* provide independent or specialist assessment where required.

***

### 51. Assurance Workflow

The standard workflow should be:

1. Identify assurance requirement.
2. Define objective.
3. Define scope.
4. Establish criteria.
5. Develop assurance plan.
6. Assign assurance resources.
7. Collect evidence.
8. Evaluate evidence.
9. Assess controls and governance requirements.
10. Identify findings.
11. Determine finding significance.
12. Develop conclusions.
13. Issue assurance report.
14. Obtain management response.
15. Track corrective actions.
16. Perform follow-up.
17. Verify remediation.
18. Close findings.
19. Report assurance results.
20. Feed lessons learned into continuous improvement.

***

### 52. Continuous Improvement

The assurance process should be improved based on:

* assurance experience;
* recurring findings;
* incidents;
* audit results;
* stakeholder feedback;
* changes in AI technology;
* regulatory developments; and
* changes in organizational risk.

***

### 53. Procedure Review

This procedure should be reviewed periodically and when material changes occur.

Review triggers may include:

* significant assurance findings;
* changes to AIGO requirements;
* regulatory developments;
* changes in assurance methodology;
* major incidents;
* changes in AI technology; and
* implementation experience.

Material changes should be versioned and approved according to applicable document governance requirements.

***

### 54. Procedure Status

**Document:** AIGO AI Assurance Procedure

**Version:** 0.1

**Status:** Draft

**Working Name:** AIGO

**Full Name:** AI Governance Operating Framework

**Document Identifier:** `AIGO-PROC-010`

**Document Type:** Operational Procedure

This procedure establishes the operational process for providing assurance over AI governance, AI systems, risks, controls, and related processes throughout the AIGO lifecycle.

***