RegWatch is an industry-grade, AI-driven regulatory intelligence platform designed to automate the discovery, classification, and monitoring of compliance obligations within the Nigerian financial ecosystem.
RegWatch employs a rigorous 4-phase lifecycle to bridge the gap between intuitive user behavior and hardened automation. This methodology ensures that every automated test is grounded in empirical reality rather than theoretical assumptions.
- Exploratory Mapping: Physical browser exploration to capture raw DOM layouts and intuitive user journeys.
- BDD Translation: Rapid conversion of explorations into strict Gherkin (Given/When/Then) syntax.
- Pre-Automation Verification: Systematic replay of BDD scenarios against the live application to secure assurance on user-centric flows before committing to code. This layer eliminates wasted effort on broken or unstable features.
- Deterministic Automation: High-fidelity Cypress implementation utilizing extensive API interception and state-injection.
The framework utilizes Gherkin Syntax as its primary behavioral source of truth. This provides:
- Universal Readability: Clear understanding for both technical engineers and business stakeholders.
- Rapid Insight: Immediate transparency into automation design and functional coverage.
- Deterministic Mapping: 1-to-1 alignment between behavioral requirements and automated assertions.
Functional prioritization is driven by a deep-level Business-Classed Risk Matrix, ensuring that engineering efforts match organizational impact.
- Tier 1 (Core - P1): Business-critical flows (e.g., Onboarding, AI Mapping, Webhook Sync). These must pass in CI/CD before any merge.
- Tier 2 (High Usage - P2): High-impact features (e.g., AI Co-Pilot, Search Filters).
- Tier 3 (Operational - P3): Supporting flows (e.g., Activity Logs, Reporting).
All tests are mapped directly to the Requirement Traceability Matrix (RTM) using unique REG-XXX identifiers, providing a continuous audit trail from requirement to verification.
To handle the inherent non-determinism of LLM-backed services and third-party webhooks, RegWatch employs several advanced engineering patterns:
- Deep API Interception: Every external AI request and webhook trigger is intercepted and replaced with high-fidelity static JSON fixtures.
- JWT State Injection: To optimize execution time, the framework bypasses UI-based login overhead by injecting JWT tokens directly into local storage via
cy.request()hooks. - Empirical Selector Governance: Selectors are never hallucinated. All locators are extracted from live browser sessions, backed by visual evidence, and stored in a centralized
selector_map.json. - Deterministic Mock Synchronization: Real-time UI updates (WebSockets/Polling) are validated by programmatically triggering mock webhook payloads during local E2E runs.
The framework integrates testAignite for advanced, quantitative throughput analysis and failure remediation.
- AI-Enhanced Root Cause Analysis (RCA): Automatically maps failure logs, investigates the technical cause, and classifies the business risk.
- Automated Fix Proposals: Generates predictive remediation suggestions for failed test schemas and DOM selectors.
- Quantitative Dashboard: A centralized portal displaying test metadata against strict performance thresholds:
- Success Rate: >90% target for E2E suites.
- Categorization Accuracy: >90% mapping precision.
- Fidelity Score: Quantitative measure of mock-to-reality alignment.
- Dashboard Load Performance: Core metrics must render in < 3 seconds.
- AI Inference Latency: Profile rescan and mapping must conclude within 30-60 seconds.
RegWatch serves as the intelligence layer within the RegTech365 Ecosystem:
- Ingestion: RegWatch identifies and classifies regulatory circulars.
- Assessment: Users execute intelligent pre-assessments to identify gaps.
- Handoff: Critical gaps are funneled to RegComply for task distribution and remediation.
- Sync: A robust webhook infrastructure ensures that task completion in RegComply is reflected instantly in the RegWatch high-level executive dashboard.