Governance of Automated Decisions: Audit, Oversight and Control for AI Decision Models Training Course

Shifting Dynamics When Judgements Are Model-Driven

• Transitioning from rule-based logic and human discretion to scored options
• The velocity and scale of automated actions
• Emerging failure modes and gaps in accountability

Inventory and Risk Categorisation of Judgements

• Creating a catalogue of automated judgements
• Assessing impact, reversibility and the parties involved
• Defining risk tiers and corresponding control expectations
• Assigning ownership for each judgement category

Accountability and Human Supervision

• Distinguishing between meaningful and nominal human review
• Setting autonomy thresholds as strategic governance decisions
• Defining escalation pathways and response timeframes
• Designating accountable owners and approval processes

Decision Logs and Explainability

• Components of a defensible decision record
• Preservation of confidence levels, available options and contextual data
• Traceability of model versions and lineage
• Communicating decisions to clients or regulators clearly

Policy and Boundary Regulation

• Managing approved option sets and change controls
• Governing thresholds through periodic reviews
• Establishing prohibited judgements and hard stops
• Implementing policy-as-code alongside model outputs

Regulatory Requirements

• Obligations under the EU AI Act for decision-making systems
• Sector-specific regulations in finance, insurance and healthcare
• Data privacy standards and rights regarding automated decisions
• Documentation and evidence of compliance

Vendor, Licence and Provenance Risks

• Comparing hosted decision APIs with open-weight models
• Understanding licence terms for derivative and embedded models
• Assessing supplier reliance and discontinuation risks
• Ensuring weight integrity and version pinning

Bias, Fairness and Outcome Surveillance

• Identifying outcome disparities across different groups
• Recognising proxy variables and the framing of options
• Monitoring signals from appeals and complaints
• Conducting regular fairness assessments

Incident Response for Incorrect Judgements

• Detecting systematic decision-making errors
• Containment strategies and suspension of autonomy
• Remediation actions and customer compensation
• Root cause analysis and updating of controls

Assurance Review of Decision-Making Layers

• Identifying evidence required by auditors
• Verifying accuracy and calibration assertions
• Sampling decisions and manual overrides
• Presenting findings to the board of directors