Teaching machines to manage exceptions: Lessons from building AI for break code, action code, and reconciliation workflows in investment operations
Exception management remains one of the highest-cost, highest-risk functions in investment operations and one of the most resistant to improvement through conventional automation. Rules-based systems have reached a structural ceiling: they cannot adapt to new break patterns, cannot prioritise exceptions by risk consequence, and cannot surface the systemic causal patterns buried in historical break data. The shift to T+1 settlement in the US, the anticipated move to T+1 in the EU,1 and the governance requirements introduced by the Digital Operational Resilience Act (DORA)2 have collectively created an inflection point at which artificial intelligence (AI)-driven exception management is no longer an experimental proposition but an operational necessity.3 This paper draws on direct experience designing and building AI capability across three specific domains — action code classification and routing, break code root cause analysis, and reconciliation exception detection and prioritisation — to provide a practitioner’s guide to what works, what fails, and what operations leaders should do to build this capability responsibly and effectively. This article is also included in The Business & Management Collection which can be accessed at https://hstalks.com/business/.
Authors
- Marcela Crossman
Publication Details
- Journal
- Journal of securities operations & custody
- Published
- 2026-11-16
- DOI
- https://doi.org/10.69554/rwvh2135
- Primary Topic
- Artificial Intelligence Applications
- Type
- article
- Field-Weighted Citation Impact
- 0.00