RECs: AI-assisted reconciliation workflows

Company

Xceptor

Role

UX Designer

Timeline

2022 – present

Team

Head of UX, Product Owners, PMs, Engineers, Sales Engineers, Delivery

[ Product screen placeholder — a real screenshot or annotated flow of the RECs recommendation-review experience goes here ]

Overview

RECs is Xceptor’s reconciliation product — one of three pillars I design across, alongside Dataset and Post-Trade. Reconciliation teams work through high volumes of transaction matches every day; RECs introduces AI-generated match recommendations into that workflow to speed the process up. My job was to make sure that speed didn’t come at the cost of trust: users needed a fast, legible way to review, question and act on what the model suggested.

The problem

Manual reconciliation is repetitive and fatigue-prone — the more matches a person reviews in a row, the easier it is to miss the one that’s wrong. Adding AI recommendations to the workflow solves the speed problem, but introduces a trust problem: if users can’t tell why the model suggested a match, they either rubber-stamp it (risky) or double-check everything anyway (defeats the point).

[Baseline metrics — e.g. average time per reconciliation task, error rate before AI assistance — to be added from product analytics.]

Process

I partnered closely with the Head of UX, Product Owners, PMs and engineers through each stage of the initiative:

North-star prototyping to align the team on where the recommendation experience should eventually go, ahead of what any single release could ship.
Internal validation with Sales Engineers and Delivery, to pressure-test the concept against real client scenarios before it reached end users.
Iterative validation with end users — reconciliation analysts — refining how confidence and reasoning were surfaced alongside each suggested match.
Release planning — working with the team to identify which parts of the north-star vision could ship incrementally.

Accessibility was a constraint throughout rather than a pass at the end: RECs sits on Xceptor’s micro-frontend design system, which I contribute to and which is assessed against WCAG 2.2.

What shipped

The result is a recommendation workflow that lets analysts move through matches quickly while still being able to see, at a glance, why the AI suggested what it did — so a “yes” is a confident yes, not a shrug.

[ Second screen placeholder — e.g. the confidence/reasoning detail view ]

Outcome

RECs now helps users complete reconciliation tasks faster and make more confident decisions on AI-generated recommendations.

[Outcome metrics — e.g. % reduction in average handling time, AI-recommendation acceptance rate, error-rate change — to be added once available.]

CEO Recognition — 2025

AI-Enabled Reconciliation Workflow Design, Xceptor