DESIGN CONCEPT · CROSS-ASSET AML COMPLIANCE · FINANCIAL CRIMES

Argos
Cross-Asset Financial Crimes Compliance Platform

One platform covering banking transactions, securities surveillance, and real estate compliance — designed for the full AML investigation lifecycle from transaction monitoring through SAR filing. Every design decision starts from the same premise: compliance tools should reduce cognitive load under time pressure, not add to it.

40
Screens Designed
4
Core Workflows
4
User Roles
BSA · FinCEN · FATF · 6AMLD
Regulations Addressed
Argos design system in Figma — 120+ compliance-specific components spanning alert cards, investigation panels, network graphs, and SAR filing workflows
Research Foundation

Financial Crimes Has Three Asset Classes.
Most Compliance Tools Only Cover One.

The defining insight from my research: AML compliance is not a single-domain problem. The same beneficial owner can move money through a bank account, trade securities through a brokerage, and purchase real estate through a shell company — and today's compliance tools treat each as a separate universe. Argos asks: what if an investigator could see all three simultaneously?

Design Concept · Research-Based Exploration — Argos is a fictional compliance platform created to explore the UX challenges specific to cross-asset financial crimes investigation. Research sources: FinCEN guidance, OCC bulletins, SEC enforcement actions, FATF recommendations; competitive analysis of Actimize (legacy), Lucinity (modern), Unit21 (API-first); domain expert consultation with Ben Brown, CPA, CFE. No NDA-protected material used.

Why this project exists: My production work (ACY Securities — 150+ compliance components, 40+ jurisdictions; Christie's — $5M–$80M UHNW property transactions) gives me direct foundations in financial compliance and real estate due diligence. This concept answers the question: what happens after the front door? When a transaction is flagged, when an alert fires, when an investigator has 22 hours of work compressed into a screen — what does that experience look like?
Research Methodology — What Informed Every Screen
Method Source / Sample Key Insight That Shaped Design
Domain Expert Consultation Ben Brown, CPA, CFE (Chartered Financial Examiner) — 3 structured interview sessions covering investigation workflows, SAR filing, and rule engine logic Investigators spend 70% of their time gathering context, not analysing it. The workspace must pre-assemble the investigation narrative — transaction timeline, entity relationships, prior SARs — before the investigator opens a case.
Production Experience 4 years at ACY Securities: 150+ regulatory components, 40+ jurisdictions, $2B+ daily settlement; Christie's: UHNW property transactions $5M–$80M Compliance isn't a feature — it's the architecture. Every screen, every data field, every workflow step must be auditable. The front-door compliance work (KYC, regulatory disclosures) I built at ACY informs the back-office investigation UX of Argos.
Competitor Audit NICE Actimize (enterprise legacy), Lucinity (modern Icelandic), Unit21 (API-first), Hummingbird (fintech-focused) — public documentation and product demos No competitor covers all three asset classes (banking + securities + real estate) in a single investigation engine. Actimize is closest but operates as separate modules with no cross-asset correlation. This is the product gap Argos occupies.
Regulatory Review BSA/AML (Bank Secrecy Act), FinCEN SAR filing requirements, FATF Mutual Evaluation recommendations, EU 6th Anti-Money Laundering Directive SAR filing has a 30-day deadline (extendable once). This time pressure shapes everything — the alert queue must be fast, the investigation workspace must be comprehensive, and the SAR form must be pre-populated from investigation findings.
Design Hypothesis Derived from all above; not validated in production The highest-leverage design intervention is the investigator workspace — where 22-hour investigations happen. If the tool can pre-assemble context, surface cross-asset connections, and auto-draft SAR narratives, the investigator's job shifts from data gathering to judgment.
The Four Users

Four Roles. One Investigation Engine.

Financial crimes compliance is a team sport. Each role has different time pressures, decision authority, and information needs. Argos gives each role a purpose-built view while sharing the same underlying case data — because an L1 analyst's dismissal becomes an L2 investigator's escalation becomes a BSA officer's regulatory filing.

L1 Analyst

Alert Triage · First Responder
  • Dismiss obvious false positives in under 2 minutes per alert
  • Escalate suspicious patterns to L2 with sufficient context
  • Process 80–120 alerts per shift without fatigue-induced errors
  • Trust the risk score to prioritise the queue correctly
Key insight: L1 analysts are speed-optimised. Keyboard shortcuts, one-click dismissals, and pre-computed risk context are the design priorities. Every additional click is a multiplied cost across 100+ daily decisions.

L2 Investigator

Deep Investigation · Case Builder
  • Build a complete investigation narrative from escalated alerts
  • Connect disparate data points across banking, securities, and real estate
  • Document findings with regulatory-grade evidence chains
  • Decide: dismiss with rationale, or recommend SAR filing
Key insight: L2 investigators are depth-optimised. The workspace must pre-assemble context — transaction timeline, entity graph, prior cases — so the investigator spends time on judgment, not data gathering.

Compliance Manager

Team Operations · Quality Assurance
  • Monitor team workload and case distribution across investigators
  • Ensure SAR filing deadlines are met across all active cases
  • Identify bottlenecks before they become regulatory violations
  • Review and approve SAR filings before submission to FinCEN
Key insight: Compliance managers are oversight-optimised. They need operational dashboards, not investigation detail — what's late, who's overloaded, where are the filing deadline risks.

BSA Officer

Regulatory Authority · Final Sign-Off
  • Maintain the institution's BSA/AML compliance program
  • Review high-risk SAR filings before FinCEN submission
  • Ensure rule engine calibration reflects current regulatory guidance
  • Report to the board on compliance program effectiveness
Key insight: BSA officers are risk-optimised. Their view shows program-level KPIs, regulatory deadlines, and audit-readiness metrics — the 30,000-foot view of institutional compliance health.
Key Design Challenges

Three Problems Worth Solving

01

The False Positive Crisis

85–95% of alerts generated by legacy transaction monitoring systems are false positives. That means an L1 analyst dismisses 85–95 out of every 100 alerts — and must stay sharp enough to catch the 5–15 that matter. The design challenge: how do you make dismissal effortless without making real threats dismissible?

The answer is risk-scored queueing (highest risk first, then age), keyboard-driven triage (d = dismiss, e = escalate, → = next), and pre-computed rule explanations that let the analyst see why the alert fired before deciding.

02

Cross-Asset Correlation

A beneficial owner purchases a $12M property through a shell company in a GTO jurisdiction, while a linked entity trades large-cap equities through a brokerage account, and a third entity wires funds through a correspondent banking relationship. Today, these are three separate investigations in three separate systems.

The challenge: how do you design an entity resolution interface that makes these connections visible, explorable, and documentable — without overwhelming the investigator with false entity matches?

03

AI Trust in a Regulated Environment

AI can accelerate investigations — suggesting connections, drafting SAR narratives, scoring confidence on entity matches. But compliance is a regulated domain: every decision must be explainable, every recommendation auditable, every automated action logged.

The challenge: how do you design an AI co-pilot that investigators trust enough to use but don't trust enough to defer to blindly? The answer: sidebar augmentation (not embedded automation), confidence scores on every suggestion, and a visible reasoning chain the investigator can accept, modify, or reject.

Domain Understanding

Financial Crimes Is Not a Trading Platform.
They Solve Opposite Problems.

Most of my production experience is in trading platforms (ACY Securities, Finlogix, LogixTrader). Before designing Argos, I had to explicitly map where those patterns apply — and where they would actively harm the compliance investigation experience.

Dimension Trading Platform (ACY / LogixTrader) Compliance Investigation (Argos)
Primary emotion Urgency, speed, competitive edge Thoroughness, accuracy, regulatory confidence
Data density Maximum — expert users want 120+ concurrent metrics High but structured — investigators need context, not raw feeds
User intent Execute — open app to trade Investigate — open app to build a case narrative
Session length Long, frequent (active traders check every few minutes) Long, deep (4–8 hour investigation sessions per case)
Time pressure Milliseconds matter — latency = lost money Days matter — 30-day SAR deadline, but thoroughness over speed
Success metric Task completion speed (order execution: 8.2s → 2.9s) Investigation quality (SAR accuracy, false positive reduction, audit readiness)
AI role Algorithmic trading — speed and automation valued Augmentation only — every AI suggestion must be explainable and auditable
What transfers Keyboard-driven interfaces, data density management, real-time data architecture, information hierarchy design

What transfers from trading platform work: keyboard-driven UI (triage shortcuts mirror trading hotkeys), data density management, real-time data architecture, cross-functional regulatory collaboration. What doesn't transfer: speed-over-accuracy mental models, automation-first AI, or the assumption that users want to act fast rather than act right.

Information Architecture

Three Asset Classes. One Investigation Engine.

Argos is built as a shared core (70%) + asset-specific modules (30%). The alert queue, case management, SAR filing, and team operations are shared. Each asset class brings its own data models, detection rules, and regulatory requirements. The cross-asset correlation engine connects entities across all three.

Banking

Banking & Payments

Transaction monitoring, AML sanctions screening, customer due diligence. 4 years of compliance infrastructure at ACY Securities — 150+ regulatory components, 40+ jurisdictions, $2B+ daily settlement volume. This is the foundation: wire transfers, correspondent banking, CTR thresholds, structuring detection.

Securities

Liquid Assets & Securities

Order book reconstruction, trading pattern analysis, market abuse detection. Informed by LogixTrader, Finlogix, and ACY's securities trading infrastructure. Pattern detection: layering, spoofing, wash trading, pump-and-dump schemes. High-frequency data ingestion with sub-second latency requirements.

Real Estate

UHNW Property Oversight

Property risk scoring, beneficial ownership tracing, GTO jurisdictional compliance. Grounded in Christie's experience with $5M–$80M asset transactions and cross-border ownership structures. Shell company detection, sanctioned entity screening, FinCEN Geographic Targeting Order monitoring.

Live Interactive Prototype

Explore the Argos App

All four workflows are fully interactive. Built in React with shadcn/ui, Tailwind, and 30+ feature components — open the prototype to navigate between the command center, alert triage, investigator workspace, and SAR filing flows.

Argos Compliance Platform — React Prototype
React · TypeScript · shadcn/ui · Tailwind · 30+ Feature Components
Command Center Alert Triage Queue Investigator Workspace SAR Filing Network Graph
Open Prototype
Design System

From User Flows to Component Library

Before any screens, the full investigation lifecycle was mapped as a user flow — from alert generation through triage, investigation, and SAR filing. The component library spans 120+ compliance-specific patterns, built to handle the data density and regulatory requirements of financial crimes investigation.

Argos user flow diagram showing the complete investigation lifecycle from alert generation through SAR filing

User Flow: Alert Generation → Triage → Investigation → SAR Filing

Argos Figma component library showing 120+ compliance-specific design patterns

Component Library: 120+ Compliance-Specific Patterns

Flow 01 · Operations Leadership
Command Center

The Operational Nerve Center — Morning Standup in One Screen

The Command Center is designed for compliance operations leadership: morning standups, shift handoffs, and real-time situational awareness. At a glance: today's alert volume, team workload distribution, risk concentration across asset classes, SAR filing pipeline, and system health. The design principle: a compliance director should know the state of the operation within 30 seconds of opening this screen.

Argos Command Center dashboard — operational overview showing alert queues, team workload, risk distribution across banking, securities, and real estate modules

Command Center — Operational Overview with Cross-Module Risk Distribution

📊
12 KPI Cards, Each Clickable

Alert volume, queue depth, SARs in review, team utilisation, filing deadlines, rule engine health — each card is a drill-down entry point. The compliance director sees the pulse; clicking reveals the granular data underneath.

🏗️
Three-Module Risk Distribution

Banking, securities, and real estate risk are shown side-by-side. If 80% of high-risk alerts are coming from the real estate module, the director can reallocate investigator bandwidth before it becomes a bottleneck.

👥
Team Workload as a First-Class Metric

Investigator utilisation isn't hidden in an HR system — it's on the main dashboard. If one investigator has 15 active cases and another has 3, the imbalance is visible before it affects filing deadlines.

🤖
AI Co-Pilot Panel for Bulk Operations

The AI sidebar surfaces operational recommendations: "3 cases approaching 30-day deadline — recommend reassignment" or "Rule 47 has generated 200+ false positives this week — consider threshold adjustment." Suggestions, not decisions.

Command Center with AI Co-Pilot panel open showing operational recommendations

AI Co-Pilot Panel — Operational Recommendations and Bulk Actions

Command Center alternative layout for mobile command room

Alternative Dashboard Layout — Optimised for Command Room Displays

Flow 02 · L1 Analyst
Alert Triage

85% of Alerts Are False Positives. Design for That Reality.

The alert triage queue is where the speed of false positive dismissal becomes the competitive edge. Argos prioritises risk-scored queueing, one-second rule engine evaluation, and keyboard-driven triage. This is where 85% of alerts are filtered out — it must be frictionless for the obvious dismissals and thorough for the genuine threats. The L1 analyst's keyboard shortcuts (d = dismiss, e = escalate, → = next) compress 8-hour investigation days into 3.

Argos Alert Queue — risk-scored, age-sorted alert list with keyboard shortcut indicators

Alert Queue — Risk-Scored Priority with Keyboard-Driven Triage

Risk Score Drives Queue Order

High-risk PEP alerts surface above medium-risk structuring patterns. Secondary sort by age prevents starvation of older medium-risk cases. The analyst's attention goes to the things that matter first.

⌨️
Keyboard-First Interaction Model

d = dismiss, e = escalate, → = next alert, ← = previous. Trading floor culture demands speed — this triage interface borrows that mental model. Mouse-optional for the entire workflow.

📋
Rule Explanation Before Decision

Every alert shows which rule fired and why. "Structuring: 4 cash deposits of $9,800 within 48 hours" — the analyst sees the pattern, not just the flag. This context is what makes a 2-minute dismissal defensible in audit.

📦
Batch Operations for Rule Corrections

When a rule generates 200 false positives, dismissing them one-by-one is not viable. Batch dismiss with a shared rationale saves hours and creates a clean audit trail for the rule adjustment that follows.

Alert detail screen showing rule explanation, transaction context, and customer history

Alert Detail — Rule Explanation and Transaction Context

Alert with transactional details inline showing wire transfer patterns

Transactional Details — Wire Transfer Pattern Analysis

Alert escalation flow showing L1 to L2 handoff with context preservation

Escalation Flow — L1 → L2 Handoff with Full Context

Alert batch actions interface for bulk dismiss with shared rationale

Batch Dismiss — Shared Rationale for Rule-Generated False Positives

Alert filtering interface with jurisdiction, rule type, and risk level filters

Advanced Filtering — Jurisdiction, Rule Type, Risk Level

Alert sorting controls showing risk, age, and investigator availability options

Sorting Controls — Risk, Age, Investigator Availability

L1 analyst personal dashboard showing shift performance metrics

L1 Analyst Personal Dashboard — Shift Performance and Queue Metrics

L1 analyst comprehensive dashboard with personal queue, performance metrics, and rule insights

L1 Analyst Workspace — Personal Queue, Performance Metrics, Rule-Level Insights

Flow 03 · L2 Investigator
Investigator Workspace

22 Hours of Investigation. One Screen to Rule Them All.

The investigator workspace is the hero of Argos — where 22-hour investigations happen. It's designed for deep dives: connecting disparate data points across banking, securities, and real estate; building narrative chains from transaction timelines; and documenting findings under the 30-day SAR deadline pressure. The AI Co-Pilot provides real-time suggestions for missing leads while maintaining full explainability and confidence calibration — because in compliance, "the AI said so" is not a valid rationale.

Argos Investigator Workspace — case detail with timeline, document explorer, and evidence chain

Case Detail — Investigation Timeline, Document Explorer, Evidence Chain

🔍
Pre-Assembled Investigation Context

When the investigator opens a case, the context is already there: the original alert, the customer profile, the transaction history, prior SARs filed on this entity, and related cases across all three asset classes. The investigator's job starts at analysis, not data gathering.

🤖
AI Co-Pilot — Suggestions, Not Decisions

Every AI suggestion shows: confidence score (0–100%), data sources used, reasoning chain, and alternative interpretations. The investigator can accept, modify, or reject. All AI interactions are logged and auditable for regulatory review.

📝
Narrative-Driven Investigation UI

Trading dashboards show numbers; compliance needs stories. The case management interface is built around a narrative timeline — "On Jan 15, Entity A wired $2.3M to Entity B, which purchased a property in a GTO jurisdiction 3 days later." This is how SARs are written, so this is how investigations should be structured.

🔗
Evidence Linking Across Asset Classes

A single investigation can reference banking transactions, securities trades, and real estate records. The workspace lets the investigator link evidence from any module into the case file — building the cross-asset narrative that no single-domain tool can produce.

Case workspace with evidence collection panel showing linked transactions and documents

Evidence Collection — Linking Transactions, Documents, and Entity Records

Case workspace showing network analysis within investigation context

In-Case Network Analysis — Entity Relationship Visualisation

Case workspace with AI Co-Pilot panel open showing suggested connections, missing leads, and confidence scores

AI Co-Pilot — Suggested Connections, Missing Leads, Confidence Scores (78% Match Confidence Shown)

Key Design Decision: AI as Sidebar, Not Embedded

The AI Co-Pilot lives in a collapsible sidebar, not embedded in the investigation flow. This is deliberate: the investigator maintains full control of the narrative. AI suggestions appear alongside the work, not inside it. The investigator can pull a suggestion into the case file with one click — or ignore it entirely. Both actions are logged. This architecture means every investigation is human-driven with AI augmentation, never AI-driven with human approval. That distinction matters to regulators.

Cross-Asset Intelligence
Network Analysis & Beneficial Ownership

The Killer Feature: See the Same Person Across Three Asset Classes

Automated beneficial ownership tracing and cross-asset correlation. A real estate transaction, a securities account, and a banking relationship are connected by the same Ultimate Beneficial Owner (UBO). Argos joins these dots in real time — and makes the connections explorable, documentable, and linkable to active investigations.

Beneficial Ownership Tracer showing entity resolution graph with ownership percentages and jurisdiction flags

Beneficial Ownership Tracer — Entity Resolution with Ownership Chains

Beneficial Ownership Tracer showing expanded network with shell company detection

Network Expansion — Following Ownership Chains Through Shell Companies

🕸️
Entity Graph, Not Just Entity List

Ownership structures are visual, not tabular. The investigator can see that Entity A owns 60% of Entity B, which owns 40% of Entity C — and that Entity C purchased a $12M property in Miami. The graph makes layered ownership structures immediately comprehensible.

🌍
Jurisdictional Risk Overlay

Each entity node shows its jurisdiction and associated risk level. A Cayman Islands holding company connected to a Panama trust connected to a US property purchase — the jurisdictional pattern is visible before the investigator reads a single document.

Cross-Asset Intelligence Dashboard showing unified customer view across banking, securities, and real estate

CAID (Cross-Asset Intelligence Dashboard) — Unified Customer View Across All Three Modules

🔗
One Customer, Three Asset Classes, One Screen

The CAID shows banking activity (wire transfers, account balances), securities activity (trading patterns, portfolio positions), and real estate activity (property ownership, transaction history) for a single entity. This is the view that doesn't exist in any competitor product — and the view that cross-asset investigations require.

⚠️
Risk Signals Aggregated, Not Siloed

A customer with low risk in banking, medium risk in securities, and high risk in real estate has a different risk profile than any single module would suggest. The CAID aggregates risk signals to produce a holistic assessment that reflects the actual threat.

Asset-Specific Module
Real Estate Compliance

$80M Properties Deserve More Than a Checkbox

Christie's UHNW experience informs this module: risk scoring for properties $5M+, sanctioned entity detection, beneficial ownership compliance, and Geographic Targeting Order (GTO) jurisdictional overlay. Every property transaction, every entity structure, every wire is cross-checked against sanctions lists, PEP databases, and adverse media.

Real estate risk assessment interface showing property value, ownership structure, and risk indicators

Property Risk Assessment — Value, Ownership Structure, Risk Indicators

Real estate beneficial ownership verification showing multi-layer entity structure

Beneficial Ownership Verification — Multi-Layer Entity Structure

Real estate financing and fund source tracking showing wire origins and intermediary banks

Fund Source Tracking — Wire Origins, Intermediary Banks

🏠
Property Risk Scoring

Every property above $5M is automatically scored based on: purchase price relative to market value, buyer entity structure complexity, jurisdiction risk, fund source transparency, and PEP/sanctions match probability. The score determines investigation priority.

🗺️
GTO Jurisdictional Overlay

FinCEN Geographic Targeting Orders require all-cash real estate transactions above thresholds in certain jurisdictions to report beneficial ownership. The GTO module automatically flags transactions in covered areas and triggers the required reporting workflow.

GTO Jurisdictional Map showing covered areas and active monitoring zones

GTO Monitoring — Jurisdictional Risk Map with Active Coverage Areas

Property Dossier View showing comprehensive property file with documents and compliance status

Property Dossier — Comprehensive Property File with Compliance Status

Property Dossier expanded view with documents, timeline, ownership history, and regulatory filings

Property Dossier Detail — Documents, Timeline, Ownership History, Regulatory Filings

Asset-Specific Module
Securities Surveillance

Order Book Reconstruction Meets Market Abuse Detection

Trade surveillance and order book reconstruction, designed with 4 years of ACY trading infrastructure plus lessons from LogixTrader and Finlogix. Pattern detection covers layering, spoofing, wash trading, and pump-and-dump schemes — with sub-second latency for high-frequency data ingestion.

Trade Surveillance Dashboard showing real-time order and execution flows with anomaly detection

TSD (Trade Surveillance Dashboard) — Real-Time Order and Execution Flows

Order Book Reconstruction showing replay of suspicious trading activity with timestamp precision

OBR (Order Book Reconstruction) — Replay of Suspicious Trading Activity

📈
What Transfers from Trading Platform Work

The TSD interface borrows directly from LogixTrader's order flow visualisation — the same real-time candlestick + volume patterns that traders use to read markets are what compliance analysts use to spot manipulation. The data is the same; the intent of the viewer is different.

⏮️
Temporal Replay as Investigation Tool

The OBR lets investigators replay order book state at any timestamp — seeing exactly what the order book looked like when a suspicious order was placed. This is the evidence that proves layering and spoofing: "at 14:32:07, Entity X placed 47 orders that were cancelled within 200ms."

Flow 04 · Regulatory Filing
SAR Filing

From Investigation to Regulatory Document — Without Starting Over

SAR (Suspicious Activity Report) filing is where investigations become regulatory documents. FinCEN deadlines are brutal: 30 days from alert to filing, extendable once. Argos automates narrative building from investigation findings, cross-references entity data, and routes through compliance review → legal approval with full audit trail. Typical SAR preparation: 22 hours. Target with Argos: 4 hours.

SAR filing form showing institution and currency transaction selection with pre-populated fields from investigation

SAR Form — Institution and Transaction Fields Pre-Populated from Investigation

📝
Auto-Populated from Investigation Timeline

Customer details, transaction summaries, and risk indicators are pulled directly from the case file. The investigator reviews and edits — they don't re-enter data they've already documented during the investigation.

30-Day Countdown as Design Constraint

The filing deadline is not buried in a settings panel — it's a visible countdown on every SAR in progress. Auto-escalation triggers at 25 days, 20 days, and 15 days, routing to compliance manager attention before the deadline becomes a regulatory violation.

✍️
Narrative Builder — Structured, Not Free-Form

SAR narratives follow a specific FinCEN structure. The builder provides section templates (Who, What, When, Where, Why, How) with suggested content from the investigation. The investigator writes in their own words, supported by the structure.

Multi-Stage Approval Workflow

Investigator → Compliance Manager review → Legal review → BSA Officer sign-off → FinCEN submission. Each stage has its own checklist and the ability to send back with comments. Full audit trail at every step.

SAR narrative builder with structured entry sections for suspicious activity description

Narrative Builder — Structured SAR Narrative with AI-Suggested Content

SAR cross-asset fields linking to related transactions and entities across modules

Cross-Asset Fields — Linking Related Transactions Across All Three Modules

SAR final review screen showing approval routing to legal and compliance sign-off

Final Review — Multi-Stage Approval Routing to Legal and BSA Officer Sign-Off

Configuration
Rule Engine

Compliance Teams Shouldn't Wait for Engineers

The Argos rule engine is configured through a visual builder: condition logic, thresholds, alert routing, and action triggers. Changes deploy in minutes, not sprints. Every rule revision is versioned and logged for audit — because when a regulator asks "why did you change Rule 47 on March 15?", the answer needs to be documented, not reconstructed.

Rule Engine visual builder showing no-code logic composition with condition blocks and threshold settings

Rule Builder — No-Code Logic Composition with Version History

🧩
Visual Logic Composition

Rules are built from condition blocks: "IF transaction amount > $10,000 AND customer risk score > 7 AND jurisdiction = GTO-covered THEN generate alert with risk level HIGH." No SQL, no engineering ticket, no 2-week wait.

📊
Impact Preview Before Deployment

Before a rule change goes live, the system shows: "This change would have generated 47 additional alerts last month" or "This change would have eliminated 200 false positives." Data-driven rule tuning, not guesswork.

Operations
Team & Role-Based Dashboards

Four Roles, Four Views, One Compliance Program

Each role gets a purpose-built dashboard: L1 analysts see their queue and shift performance; investigators see their active cases and deadlines; compliance managers see team workload and filing pipeline; BSA officers see program-level KPIs and regulatory deadlines.

Team Lead dashboard showing investigator workload distribution and case assignment status

Team Lead Dashboard — Investigator Workload and Case Distribution

Team dashboard with AI workload balancing showing recommended case reassignments

AI Workload Balancing — Recommended Case Reassignments

BSA Officer dashboard showing program-level compliance KPIs, regulatory deadlines, and audit readiness

BSA Officer Dashboard — Program-Level Compliance Health and Regulatory Deadlines

Supporting Screens
Watchlist, Command Palette & Settings

The Details That Signal Production Readiness

Three supporting screens that round out the platform: sanctions and PEP watchlist management, keyboard-driven command palette (Cmd+K), and system configuration for API integrations, user preferences, and data source management.

Watchlist Management showing sanctions, PEP, and adverse media monitoring

Watchlist Management — Sanctions, PEP, Adverse Media

Command Palette showing keyboard-driven navigation with Cmd+K

Command Palette — Keyboard-Driven Navigation (Cmd+K)

Settings screen showing user preferences, API keys, and integration configuration

Settings — User Preferences, API Keys, Integration Configuration

Design Decisions

Every Trade-Off, Documented

Design is choosing. These are the major architectural and interaction decisions that shaped Argos — each with the rejected alternative and the reasoning for the chosen approach.

Multi-Asset vs Single-Asset Architecture
Rejected
Single-asset platform (banking only)
Simpler to build, faster MVP. But real compliance is cross-asset: the same person owning a property, trading securities, and moving money. Segregation creates blind spots that criminals exploit.
Chosen
Unified cross-asset investigation engine
Longer build, but Argos becomes the source of truth. A single customer can be understood across all activities. This is the defensible product moat — competitors are still reconciling three separate systems.
AI Co-Pilot Integration Model
Rejected
Embedded AI (autonomous decision-making)
Full automation sounds efficient, but compliance requires explainability. If the system approves a SAR filing, regulators want to know why. Black-box AI is a liability, not an asset.
Chosen
AI as sidebar co-pilot (augmentation model)
Suggestions, not decisions. Every AI output shows confidence score, source data, and reasoning. Investigator accepts/rejects. All AI actions are logged. This is the regulatory-friendly AI approach.
Alert Queue Sorting Strategy
Rejected
Chronological (oldest first)
Fair, but inefficient. Some alerts are routine, others are serious. FIFO means a high-risk PEP alert waits 3 days behind suspicious-deposit noise.
Chosen
Risk-scored (highest first, then age)
Investigator attention goes to the things that matter. Secondary sort by age prevents starvation of older medium-risk cases. Fast-path for obvious dismissals via L1 rules.
Command Center Granularity
Rejected
Minimalist dashboard (3 KPIs)
Less is more? Not for compliance operations. The director needs situational awareness: alert inflow, SARs in review, team bottlenecks, filing deadlines.
Chosen
Rich operational dashboard with drill-down
12 major KPI cards, each clickable to underlying data. Compliance director sees the top-level pulse; drilling reveals granular details. Information density is the feature, not the problem.
Regulatory Framework

Every Screen Traces Back to a Compliance Mandate

Argos design is anchored to regulatory requirements. This isn't decorative domain knowledge — it's the architecture. Every workflow, every data field, every approval step exists because a regulation requires it.

BSA

Bank Secrecy Act / AML

Transaction monitoring, Currency Transaction Report (CTR) filing, and SAR reporting. The BSA's five pillars (internal controls, BSA officer, training, independent testing, customer due diligence) are embedded in the team operations and BSA officer dashboard. Alert queue and investigation workspace directly support these obligations.

FinCEN

FinCEN SAR Requirements

30-day filing deadline (extendable once to 60 days). SAR filing workflow enforces this timeline with visible countdown, auto-escalation triggers, and multi-stage approval routing. The narrative builder follows FinCEN's structured format: Who, What, When, Where, Why, How.

FATF

FATF Mutual Evaluation

Financial Action Task Force recommendations on beneficial ownership transparency, sanctions screening, and cross-border reporting. Incorporated into the CAID (Cross-Asset Intelligence Dashboard) and GTO monitoring modules.

6AMLD

EU 6th Anti-Money Laundering Directive

Enhanced customer due diligence for high-risk entities, real estate transparency register requirements, and PEP screening obligations. The Real Estate module is built to 6AMLD beneficial ownership standards.

Research & Methodology

What Was Researched. What Was Assumed.

Transparency about what's real and what's hypothetical is a design decision in itself. Concept projects that pretend everything is validated lose credibility with senior reviewers. Here's the honest breakdown.

Domain Expert Consultation — Ben Brown, CPA, CFE (Chartered Financial Examiner), reviewed all compliance workflows, rule engine logic, and SAR filing automation across 3 structured interview sessions. His feedback strengthened investigator workflow credibility and regulatory realism throughout the design.
Category What Was Researched Source
BSA/AML regulatory landscape Filing requirements, enforcement actions, compliance program structure FinCEN guidance, OCC bulletins, SEC enforcement actions
Transaction monitoring at scale Rule engine architecture, false positive patterns, alert queue design 4 years at ACY Securities: 150+ rules, 40+ jurisdictions
UHNW real estate compliance Beneficial ownership structures, GTO requirements, cross-border deals Christie's: $5M–$80M property transactions
Securities surveillance Order book reconstruction, market abuse pattern detection LogixTrader, Finlogix architectures; ACY trading infrastructure
Competitive landscape Feature gaps, UI patterns, market positioning Actimize (legacy), Lucinity (modern), Unit21 (API-first)
Assumption Why It Matters What Production Would Require
Real estate risk scoring Model assumes access to property prices and ownership registries Partnerships with data providers (CoreLogic, public registries)
Cross-asset entity matching Assumes probabilistic UBO matching — name alone is not sufficient Identity reconciliation engine (Dun & Bradstreet, graph databases)
AI confidence calibration Co-Pilot confidence scores are placeholder values Bayesian updating based on investigator accept/reject decisions
Latency assumptions Alert queue sorting assumes rule engine evaluation < 500ms Sub-100ms latency engineering for securities module specifically
Metrics Framework

If Argos Were Shipped, These Are the KPIs

Concept projects that can't articulate their success metrics aren't thinking about outcomes. These are the six KPIs that would determine whether Argos is working — each with a specific target and measurement methodology.

<8h

Alert-to-Decision Time

From baseline 22 hours. Measured end-to-end: alert fired → investigator decision (dismiss/escalate/SAR). Tracked per role and rule category. The single most important operational metric.

<2m

False Positive Dismissal

For obvious false positives. Measured as time from alert queue view to dismissal action. Keyboard shortcuts are critical to hitting this target. If it takes longer than 2 minutes, the rule context isn't clear enough.

50%

SAR Preparation Reduction

From 22 hours to 11 hours (legal review still required as human-in-the-loop). Auto-populated narratives and cross-asset linking save the bulk of data-gathering time.

<45

NASA-TLX Cognitive Load

Post-task workload assessment quarterly. Argos should reduce mental effort, not increase it. If investigators are more cognitively loaded with the tool than without it, the design has failed.

60–70%

AI Co-Pilot Acceptance Rate

Adoption of AI suggestions. Too low = Co-Pilot not trusted. Too high = might be automating judgment that should remain human. The sweet spot indicates useful augmentation without over-reliance.

<5m

Rule Maintenance Velocity

Time to deploy a rule change. No-code builder should enable compliance teams to iterate without engineering tickets. Currently most organisations wait 2–4 weeks for rule changes.

Reflection

What Transferred. What Was New. What I'd Iterate.

The Most Ambitious Case Study I've Created

Argos represents 4 years of compliance domain knowledge compressed into a unified platform design. It's not shipped — but every decision is defensible and grounded in real regulatory requirements, investigator workflows, and domain expert feedback. The cross-asset architecture is the product thesis: no competitor covers banking, securities, and real estate in a single investigation engine. That gap is real, and this design shows what filling it looks like.

Category Detail
What transferred from trading Real-time data architecture (LogixTrader/Finlogix → securities module), keyboard-driven UI patterns (trading hotkeys → triage shortcuts), data density and visual hierarchy management
What was new Narrative-driven investigation UI (trading shows numbers; compliance needs stories), regulatory compliance as design constraint (every screen must be audit-friendly), cross-domain unification (reconciling three asset-class data models)
What I'd iterate Entity resolution UI (UBO matching is probabilistic — needs investigator-guided matching interface), AI confidence calibration (needs Bayesian updating from investigator feedback), mobile responsiveness (compliance teams need on-call access), third-party data integration patterns