AI-Driven Compliance Monitoring, Governance and the Future of Integrated Oversight
- Operational and cultural barriers remain the biggest challenge in AI-driven continuous monitoring.
- AI delivers immediate value in surveillance, horizon scanning and regulatory reporting.
- Human judgement remains essential for complex compliance decisions.
- Governance, auditability and explainability are critical to trusted AI outcomes.
- AI will enable more integrated oversight models across risk, compliance and assurance functions.
- The future of governance is human-led, AI-augmented oversight.
As organisations increasingly deploy AI across compliance and risk functions, moving from periodic reviews to continuous monitoring presents both opportunities and challenges. In this Q&A, Suresh Sankaran, Prudential Regulation Lead at NatWest Group, shares his perspectives on AI-enabled oversight, evolving regulatory expectations, governance frameworks and the balance between automation and human judgement.
1. Moving from periodic reviews to continuous monitoring is a major shift for many organisations. What are the biggest operational and cultural barriers you've seen when implementing AI-driven continuous oversight, and how can firms overcome them?
The biggest challenge is rarely the
technology itself. It is the operating model and the mindset shift.
Traditionally, compliance and
assurance functions have been built around periodic reviews, sample testing and
retrospective analysis. Continuous monitoring challenges that model because it
creates a near real-time view of risk and control effectiveness.
Key operational barriers include
data quality and integration challenges, fragmented control environments,
unclear ownership of alerts and remediation activities, and alert fatigue.
Cultural barriers include concerns around AI replacing professional judgement,
resistance to increased transparency and viewing monitoring solely as a
compliance responsibility rather than a business responsibility.
Successful organisations position AI
as an augmentation capability, supported by strong data governance, clear
accountability and a focus on identifying emerging risks earlier.
2. As regulatory expectations continue to evolve, where do you see AI delivering the most immediate value within compliance surveillance and regulatory monitoring, and which use cases are still not mature enough to rely on?
The most mature use cases are those
involving large volumes of information that require efficient analysis and
prioritisation.
Immediate value can be found in:
- Regulatory horizon scanning and change monitoring.
- Policy and regulatory obligation mapping.
- Transaction and communication surveillance.
- Anomaly detection.
- Evidence gathering and controls testing support.
- Regulatory reporting quality assurance.
Areas requiring complex contextual
judgement remain less mature, including conduct assessments, interpretation of
regulatory intent, disciplinary decisions, unusual control evaluations and
final assurance opinions. AI can highlight potential issues, but humans should
remain responsible for final decisions.
3. When AI is used for controls testing and assurance activities, how should organisations measure effectiveness and ensure the outputs are reliable enough to support critical risk and compliance decisions?
Many organisations focus too heavily
on efficiency metrics, but effectiveness should be measured more broadly.
Key dimensions include:
Accuracy
- False positive rates.
- False negative rates.
- Precision and recall.
Coverage
- Percentage of activity monitored.
- Degree of automation achieved.
- Extension beyond traditional sample-based testing.
Outcome Quality
- Earlier risk identification.
- Reduced regulatory findings.
- Improved remediation outcomes.
Governance
- Explainability.
- Auditability.
- Reproducibility.
- Independent validation.
AI outputs should be regularly
validated, challenged and back-tested, with standards comparable to established
model risk management principles.
4. Looking ahead three to five years, do you think compliance, risk and assurance functions will remain separate disciplines, or could AI enable a more integrated oversight model across the entire organisation?
AI will drive greater integration
between these functions, although not complete convergence.
A shared risk intelligence layer can
continuously monitor risks, test controls, gather evidence and generate
management information from a single source of truth.
However, the disciplines will remain
distinct because risk management, compliance and assurance each fulfil
different governance objectives. AI may unify workflows and information, but
independent oversight remains essential.
5. As AI becomes more capable of identifying risks, testing controls and recommending actions, where should organisations draw the line between automated decision-making and human judgement, and do you expect that boundary to change over time?
The guiding principle should be
clear: AI can inform decisions, but accountability should remain with humans.
AI excels at processing information
at scale, identifying patterns, continuous monitoring and recommending actions.
Human involvement remains crucial for contextual assessment, ethical judgement,
materiality considerations and ultimate accountability.
While technological maturity and
regulatory frameworks will continue to evolve, highly regulated sectors are
unlikely to move towards fully autonomous compliance and risk decisions in the
foreseeable future.
The future is likely to be
characterised by human-led, AI-augmented oversight that combines machine
efficiency with human expertise and accountability.
The real opportunity is not to
automate compliance. It is to move from discovering control failures after the
event to identifying emerging risks while there is still time to act. AI
changes the speed and scale of oversight, but good governance still depends on
human accountability.