- Published On Oct 28, 2026 at 05:24 PM IST
AI-generated reports could also improve consistency by using the same transaction analysis periods and report structures across cases, with claims linked to their source data for review by quality teams, auditors and regulators.
However, McKinsey stressed that human oversight remains essential. AI agents can analyse evidence and make recommendations, but people must assess findings, address gaps in information and retain responsibility for decisions.
“AI augments our work, but all judgments remain with our team,” Hans-Georg Beyer, Commerzbank’s chief credit risk officer and former group chief compliance officer, said in the report.
McKinsey said banks seeking to scale the technology should build systems with quality checks, clear risk controls and traceable decisions. It noted that 30-50% of an agent network may need to be dedicated to quality assurance and traceability.
The report also said the same approach could be extended beyond transaction monitoring to fraud detection, know-your-customer checks and regulatory compliance. One large bank used a multi-agent system based on similar principles to prepare corporate credit risk assessments, halving the time credit officers spent on analysis.
McKinsey said banks should begin with a focused use case, demonstrate its value and then expand to wider compliance activities. It cautioned that successful adoption would depend not only on technology and data access but also on staff skills, continued collaboration between compliance experts and data scientists, and robust governance.
The report said banks would need to monitor deployed AI systems closely, test their reliability and make continuous improvements as risks and business requirements evolve.
- Published On Oct 28, 2026 at 05:24 PM IST
Join the community of 2M+ industry professionals.
Subscribe to Newsletter to get latest insights & analysis in your inbox.
All about industry right on your smartphone!
- Download the App and get the Realtime updates and Save your favourite articles.