Why this matters
ICA’s September 2026 programme asks where human judgement still matters in compliance. The answer should not be framed as humans versus machines. The real question is which decisions require accountable judgement because the consequences cannot safely be delegated to automated output.
Good uses of automation
Automation is effective for data matching, alert prioritisation, document extraction, pattern identification and repetitive workflow. These are areas where machines can reduce manual burden while preserving human oversight.
Decisions requiring judgement
Suspicious activity escalation, customer exit, interpretation of ambiguous sanctions ownership, regulatory breach assessment, high-risk customer acceptance and material policy exceptions often require contextual judgement. These decisions can affect rights, reputations and regulatory obligations.
Designing human review
Reviewers need enough information to challenge the system, not just approve its recommendation. Interfaces should display relevant evidence, uncertainty and reasons rather than hiding complexity behind a single score.
Accountability
Every material AI-assisted decision should have an accountable human owner. Governance should define which decisions may be automated, which require mandatory review and which may never be made solely by an AI system.
Conclusion
Human oversight is valuable when it changes the decision process. A human who merely clicks approve is not a control. Effective governance preserves human accountability where judgement, proportionality and ethical reasoning remain essential.
Practical actions for compliance leaders
- Assign clear ownership and document decision rights.
- Test control effectiveness using actual case outcomes rather than policy completion alone.
- Escalate ambiguous or high-consequence cases to appropriately skilled reviewers.
- Track exceptions, overrides, quality assurance findings and recurring root causes.
- Ensure board and senior management reporting focuses on risk and control effectiveness, not only volumes.
- Maintain evidence showing how the firm reached material compliance decisions.
Related reading
See Model Risk, Regulatory Compliance and Governance, Risk and Compliance.
About the author
Jonas Adam Mohamed Osman Abdelghafour writes about governance, risk and compliance, anti-money laundering, financial crime prevention, sanctions, customer due diligence, AI governance, model risk and regulatory risk. His work focuses on practical control design, risk-based decision-making and the interaction between regulation, technology and financial services. See About and Qualifications.
Source and editorial context
This article is original analysis informed by a current compliance theme highlighted by the International Compliance Association. The ICA is cited as an editorial/current-issue source; no affiliation with or endorsement by the ICA is claimed or implied.
Frequently asked questions
Why this matters?
ICA’s September 2026 programme asks where human judgement still matters in compliance. The answer should not be framed as humans versus machines. The real question is which decisions require accountable judgement because the consequences cannot safely be delegated to automated output.
What should risk leaders know about good uses of automation?
Automation is effective for data matching, alert prioritisation, document extraction, pattern identification and repetitive workflow. These are areas where machines can reduce manual burden while preserving human oversight.
What should risk leaders know about decisions requiring judgement?
Suspicious activity escalation, customer exit, interpretation of ambiguous sanctions ownership, regulatory breach assessment, high-risk customer acceptance and material policy exceptions often require contextual judgement. These decisions can affect rights, reputations and regulatory obligations.
What should risk leaders know about designing human review?
Reviewers need enough information to challenge the system, not just approve its recommendation. Interfaces should display relevant evidence, uncertainty and reasons rather than hiding complexity behind a single score.
What should risk leaders know about accountability?
Every material AI-assisted decision should have an accountable human owner. Governance should define which decisions may be automated, which require mandatory review and which may never be made solely by an AI system.