Skip to content

01 — Introduction to Regnify

Regnify is a cloud-based Software-as-a-Service (SaaS) platform that helps financial institutions in Singapore digitise and automate the licensing of their representatives under the Monetary Authority of Singapore (MAS) regulatory framework. It covers the full lifecycle — from a representative's initial declaration through multi-party internal approval to final MAS submission — replacing disconnected email chains, spreadsheets, and paper forms with a single, auditable digital workflow.


1. The Compliance Challenge

Before Regnify, onboarding a licensed representative at a financial institution involved multiple manual steps spread across several teams:

  • Form 3A (Appointment of Representative) required a representative to hand-fill a nine-section paper or PDF form, attach supporting documents, and pass it physically or by email to HR and Compliance teams.
  • Approval chains of up to six sequential approvers had no centralised tracking; a misplaced email caused days of delay with no visibility into the blockage.
  • MAS submission — once internal approval was complete — required Compliance to rekey data into the MAS portal, creating a second source of entry errors and version mismatch.
  • Ongoing obligations such as Continuing Professional Development (CPD) hours under FAA-N26, CMFAS examination status, annual attestations, and incident disclosures were tracked separately in spreadsheets with no linkage to declaration records.
  • Audit trail was fragmented across email inboxes, making regulatory examination preparation expensive.

The result was slow onboarding (often weeks), high error rates, and compliance risk from missed deadlines or inconsistent records.


2. How Regnify Solves It

Regnify reduces the onboarding cycle to a guided digital flow:

  1. Digitise — The representative completes Form 3A as a single scrollable page with 14 MAS-specified sections and a collapsible Table of Contents sidebar. The form validates all required fields on submission, requires at least one supporting document upload, and auto-saves progress as a draft every 30 seconds.
  2. Automate — On submission, Regnify automatically routes the declaration through the organisation's configured approval chain (up to six steps). Each approver receives a notification, can approve or send back with comments, and the system tracks every action with a timestamp.
  3. Submit — Once internal approval is complete, the Compliance team uses the MAS Console to review a read-only summary and initiate MAS submission directly from the platform — no re-entry of data.
  4. Monitor — After submission, MAS review status updates flow back into the platform. All historical declarations, workflow decisions, CPD records, attestations, and audit logs remain searchable at any time.

3. What Makes Regnify Unique: AI at Every Step

Regnify embeds an AI compliance assistant throughout the platform — not as a generic chatbot, but as a specialist with deep knowledge of MAS regulations and write access to the platform itself.

200-Document MAS Regulatory Knowledge Base

The AI assistant is grounded in a curated knowledge base of 200 MAS regulatory documents organised across 17 categories (Acts, Notices, Circulars, Guidelines, FAQs, and more). When a user asks a regulatory question, the assistant retrieves the relevant document sections and cites them by name — it does not guess from training data alone.

14 Specialised AI Tools

The assistant does not just answer questions. It can take structured actions through 14 built-in tools:

Tool What it does
search_mas_knowledge Full-text search across all 200 regulatory documents
read_document Fetch the full content of a specific regulatory document or section
get_declaration Retrieve a declaration's current state (with NRIC, date of birth, and phone automatically masked)
get_workflow_state Inspect where a declaration is in its approval chain
validate_nric Verify a Singapore NRIC or FIN number using the official checksum algorithm
spawn_agent Fan out multiple research tasks in parallel for complex queries
analyze_cert_gap Run a CMFAS certification gap analysis against the required modules for a set of regulated activities, and persist results
get_rep_register Retrieve all representatives under an organisation with their Red/Amber/Green compliance health flag
get_cpd_status Retrieve a representative's CPD-hour totals for a given cycle year
draft_cpd_reminder Compose a personalised CPD reminder email (drafted for human review before sending)
launch_attestation_cycle Create an attestation cycle for a set of representatives under an organisation
detect_attestation_anomaly Cross-reference a representative's attestation answers against their prior declarations and audit history to surface inconsistencies
create_incident_declaration File a new incident disclosure declaration on behalf of a representative
generate_passport Issue or rotate a representative's regulatory passport export token

7 Role-Specific Skill Playbooks

The assistant is configured with seven skill playbooks spanning five specialised domains: regulatory guidance, workflow management, document processing, certification analysis, and platform operations. Each playbook shapes the assistant's behaviour to match the regulatory and operational context of the current conversation.

Real Write Actions — Not Just Advice

Most AI assistants can only describe what to do. Regnify's AI can actually do it: launch an attestation cycle, file an incident declaration, generate a passport export, or draft a CPD reminder — all with a full audit trail linking every action to the AI agent that initiated it.

PII Protection Built In

When the AI assistant retrieves a declaration, it never sees the raw personal identifiers. At the tool boundary, NRIC and passport numbers are masked to the first and last character (for example, S****A), phone numbers show only the last four digits, and dates of birth show only the year. This masking is deterministic and happens before any data reaches the language model.

Proactive Intelligence

Regnify does not wait for you to ask. The notification system distinguishes AI-initiated alerts (marked with an "AI" badge) from human-authored ones. When an AI agent detects a deadline risk, a CPD shortfall, or an attestation anomaly, it sends a proactive notification so the responsible person can act before a deadline is missed.

Full Audit Trail for Every AI Action

Every tool call the AI assistant makes is logged in the platform audit trail with the same record type used for human actions. Compliance teams can search, filter, and export AI-initiated actions for regulatory examination purposes.


4. User Roles Overview

Regnify has six roles. Each role controls which pages a user can access and what actions they can perform.

Role Display name Primary responsibility
SYS_ADMIN System Administrator Platform-wide administration: all organisations, all declarations, user management, audit logs
ORG_ADMIN Organisation Administrator Organisation-scoped administration: declarations, workflows, users within the organisation
HR_ADMIN Human Resources Approval queue management; workflow template configuration
COMP Compliance MAS Console access; attestation cycle management; representative register and health monitoring
FI_USER Financial Institution User Declaration creation and submission; Fit and Proper assessments
REP_USER Representative Self-service representative portal: passport, attestations, CPD log, incident disclosure

A user can hold only one role. The platform renders a different navigation menu and dashboard for each role based on the permissions assigned to it.


5. Platform Architecture

Regnify is a three-tier web application:

  • React frontend — Browser-based application built with React, TypeScript, and Tailwind CSS. Accessed at https://regnify.regnify.dev.
  • NestJS GraphQL backend — Handles all declarative data: user management, declarations, workflow instances, organisations, audit logs, and representative profiles. Uses PostgreSQL for storage.
  • FastAPI AI backend — Handles all AI operations: the compliance assistant, document processing, PDF and DOCX export, email workers, and background processing via RabbitMQ. The AI assistant is powered by a language model accessed through the Cerebras API.
  • Keycloak — Handles authentication and single sign-on (SSO) using the OpenID Connect protocol with PKCE code flow. All sessions are token-based; the platform never stores passwords.

All infrastructure runs on Kubernetes. There is no data stored outside Singapore.


Related sections: Getting Started | Dashboard by Role