Flagship product case study
Designing a modular anti-financial-crime product system — and the human-owned, AI-assisted operating model used to shape, implement and validate it.
Financial-crime operations combine large queues, complex rules, fragmented evidence, regulated decisions and high expectations for auditability. AFC Suite treats that challenge as a product system, not a collection of disconnected screens.
The suite brings focused runtime experiences together on shared foundations for identity, tenancy, product access, workflow, configuration governance, provider patterns and audit. Each product stream owns its domain language and operating decisions while reusing consistent platform controls.
My contribution combined product and AI strategy leadership, workflow and UX direction, architecture shaping and delivery governance. It did not replace engineering accountability.
Operations workbench, customer and risk views, screening, alerts, cases, payment screening, transaction monitoring and governed reporting.
Implemented synthetic demonstration surfaces shown belowTrade-package intake, document processing, TBML rules, compliance findings, alerts, cases, reporting and governed product configuration.
Represented through portfolio-safe product and architecture visualsEntity and person search, ownership tracing, AI-assisted research, report generation, workbench review actions and product administration.
Represented through portfolio-safe product and architecture visualsA product leader and software engineer used ChatGPT, Codex and GitHub Copilot within bounded roles, human review, automated testing, browser validation and evidence-based release gates.
“AI-native” is not a licence to remove judgement. The delivery model distinguishes exploration from implementation, and implementation from acceptance. AI outputs are reviewed against active specifications, security expectations, tests and browser evidence. Failed patterns feed back into rules, documentation and regression controls.
The current implementation uses a modular-monolith architecture: product modules keep their own domain models and decisions while consuming shared platform contracts.
The portfolio emphasises decision architecture rather than infrastructure inventory: where product autonomy is useful, where shared controls are mandatory, and how workflow, identity, audit and provider boundaries remain governable across modules.
The interface direction is built around a simple principle: dense operations need fast scanning, clear context and deliberate decision points.
Use enterprise data grids for queues and high-volume operational work, with clear filters, density controls and prioritisation.
Provide fast context without forcing users to abandon the queue, preserving place and reducing navigation cost.
Use cards for recommended actions, exceptions and compact explanatory content — not as a replacement for operational data structures.
Keep high-frequency, high-consequence actions visible while making authority, status and validation requirements explicit.
The AFC design system is a hybrid, specification-led and code-led practice. Written UX decisions, reusable front-end components and browser evidence form the implementation source of truth. This portfolio does not claim a Figma-led library where one was not the governing artefact.
Dense operational work uses a reusable grid-and-workspace pattern with filters, tabs, status context, density controls and guided actions.
Evidence: implemented synthetic screen
Operational signals are separated from recommended work so the user can scan state, understand urgency and act without losing context.
Evidence: reusable information hierarchy
Tabs organise depth, contextual controls preserve task focus and high-consequence actions remain visible without overwhelming the page.
Evidence: shared interaction pattern
Loading, empty, failure and permission states are designed as part of the component contract rather than treated as implementation afterthoughts.
Evidence: resilient workflow designThese images are selected to show information hierarchy, workflow controls, reusable patterns and operational states. Counts are interface state, not customer adoption, production volume or performance evidence.
This is a portfolio case study, not a sales assurance pack, regulatory attestation or production-readiness statement.
This case study reflects the way I work: model the system, make decisions explicit, build reusable patterns and require evidence before acceptance.