The UI · No-code

Resolve the 2% — fast, auditable, no code

The deterministic engine clears ~98% of addresses automatically. The Workbench routes the rest through a fast, auditable queue, with an AI-assisted layer for genuinely ambiguous cases.

See the Workbench Open the platform overview

The ARS Exceptions Workbench is the no-code web application where operations, engineering and compliance teams resolve the small share of addresses the engine cannot auto-correct. The deterministic engine clears ~98% automatically; the Workbench routes the rest through a fast, auditable queue.

The two-tier model

>99% straight-through resolution

~98%

Deterministic, sub-50ms P95

~2%

Agentic Workbench, <2s

>99%

Combined straight-through

The combined outcome is straight-through resolution above 99% while preserving the deterministic audit trail regulators require.

Three lenses

Role-tuned dashboards

One web application, one data set, three lenses.

Operations

Live oversight of throughput, STP rate, exceptions and SLA, with a 2-second-refresh processing stream and SLA-at-risk indicators.

Management

Counterparty and currency quality analytics, blended cost per transaction, and SLA compliance by scheme.

Analyzer

Deep STP and repair-rate trends, scheme breakdowns and repair-outcome attribution for continuous-improvement programmes.

Live oversight

Operations dashboard

Operations · Processing streamLIVE · 2s refresh

98.4%

STP rate

1,284

TPS

In queue

42ms

P95 latency

The exception queue

Fast, auditable workflow

Verify / Processed tabs with rich filtering and global search.
One-click resolution with the Accept → Verify → Review pipeline.
Field-by-field provenance — source status (Matched / Derived), confidence scores, and the reason codes behind every decision.
SLA-at-risk indicators so ageing items are caught before they become incidents.

The result

A measurable drop in cost-per-exception — from $15–40 toward under $5 once correction is embedded.

For the ambiguous 2%

Agentic Workbench & ALS

Agentic Workbench

Proposes a structured resolution and surfaces its reasoning, with an optional human-in-the-loop step for sign-off. Returns results in under two seconds and never bypasses the audit trail.

Autonomous Learning System

Continuously self-tests resolution quality and updates parsing rules with no manual re-training. Accuracy rises from 85–90% to 95%+ within the first quarter — no engineering intervention.

Built-in tools

Playground

A real-time, no-code sandbox that accepts free-text, database records, ISO 20022 XML or SWIFT MT and returns a parsed result with four views.

ARS Playground · try a free-text address

Input address (free-text, hybrid or structured)

Fields ISO XML Database row Explanation

File Uploads

Drag-and-drop CSV/XLSX for bulk remediation, with row-level evidence for governance.

Benchmark Arena

Compare resolution outcomes across configurations before they reach production.

Audit & retention

The evidence base for examination

365-day retention, per-tenant

Every processing event persists in the ARS database, keyed by request and file IDs, scoped per tenant with no cross-tenant visibility. File backups carry a 365-day retention default.

FAQ

Frequently asked questions

No. The Workbench is a no-code web application; operations and compliance teams resolve exceptions through a queue with one-click actions.

The deterministic engine clears about 98% automatically. The remaining ~2% route to the Workbench, where the Agentic layer plus optional human review lifts the combined rate above 99%.

It applies AI reasoning only to the small ambiguous remainder, with human-in-the-loop review, and never replaces the deterministic, fully audited core.

Yes. Every decision is logged with provenance and reason codes, scoped per tenant, with a 365-day retention default.