From 6,000 breaks to under 2,000: productizing exception management

A reconciliation operation drowning in manual breaks — rebuilt discovery-to-delivery like a product: mapped workflows, wrote the requirements, led UAT, shipped playbooks, and measured everything.

The problem

Securities lending settlement runs on reconciliation — matching what should have happened against what did. When I owned these workflows, the picture was familiar to anyone in ops: stuck, suspense, and compare-break queues cleared by hand, exception volumes that never really came down, and aged breaks sitting 5+ days because root causes kept recurring. Every break was being worked; almost none were being prevented.

The approach

I treated it as a product problem, not a staffing problem:

• Discovery first. Mapped the stuck, suspense, and compare-break workflows end to end across global markets — where the handoffs broke, which steps added no value, where the same break recurred under different names.
• Redesigned the workflow, not just the queue. Led a discovery-to-delivery redesign that eliminated 30% of manual steps and cut cycle times 20–25%.
• Wrote the requirements like a PM. Authored business requirements for Aladdin lending-module enhancements and led UAT — acting as the operational product lead between business users and engineering. The result: fewer manual touchpoints, better straight-through-processing rates, and a 20% reduction in Aladdin loan/return exceptions.
• Attacked recurrence with root-cause analysis. Drove a metrics-based exception-reduction program with tightened control frameworks — recurring issue categories fell ~75% and ~49%, quantity-related breaks dropped from 54 to 20.
• Made the fix durable. Published standardized exception-management playbooks (Custody Breaks, Cash & Security Holdovers, Quantity Compare Breaks) — repeat exceptions fell 15–20% and aged-break duration dropped from 5+ days to ≤2 days.
• Brought AI into the workflow. Embedded generative AI into reconciliation prep, exception triage, and documentation — building reusable prompt libraries and agent workflows that became the default starting point for high-volume tasks, cutting task completion time 25–30%.

The results

• Failing settlement items: 6,000+ → under 2,000
• Holdover items: 153 → 62
• Quantity-related breaks: 54 → 20
• Manual steps eliminated: 30%; cycle times down 20–25%
• Aladdin loan/return exceptions down 20%
• Zero-fail go-lives across platform migrations, real-time collateral management, and custodian consolidation workstreams
• Mentored 3 junior colleagues on operational continuity and knowledge transfer

What this shows a hiring manager

Most candidates can describe a process. Fewer can show they productized one: defined the problem with data, wrote requirements engineers could build from, validated with UAT, drove adoption with playbooks, and proved the outcome in metrics. That's the skill set I bring to Product Operations and AI Operations roles — and it's the same discipline behind the tools on this site.