RegulaPilot
AI-powered compliance workflow system for structured decision-making
Problem
Compliance teams operate in environments where large volumes of dense regulatory text must be translated into actionable decisions. Existing tools focus on summarisation, but fail to convert information into structured workflows — leaving teams to manually extract obligations, assign ownership, and manage risk under time pressure.
Approach
Workflow-first design: instead of summarising text, the system structures outputs into operational components — obligations, risks, owners, deadlines, priorities, and source references. The focus was on designing the correct product shape first, with a clean UI and deterministic schema, while stubbing the AI layer to prioritise speed of execution and iteration.
Architecture
Frontend-driven MVP architecture designed for rapid iteration, with a clear separation between input handling, processing logic, and structured output rendering. The system is intentionally built to evolve into a full LLM-backed pipeline with schema validation.
Input Layer
User pastes regulatory or compliance text (FCA, AML/KYC, policies, contracts)
Processing Layer (Stubbed)
Simulated AI pipeline returning structured outputs (planned: LLM with JSON schema enforcement)
Schema Layer
Defines structured output: obligations → owner → deadline → priority → source quote
Presentation Layer
Dashboard UI displaying obligations, risks, and tasks in a workflow-oriented format
Future Data Layer
Planned persistence layer for audit trail, task tracking, and regulatory reporting
Key Engineering Decisions
Workflow over summarisation
Most AI tools generate summaries. This system structures outputs into actionable workflows, making it usable in real operational environments rather than just informative.
Speed-first MVP development
Built in 24 hours to prioritise product validation over completeness. The AI layer is stubbed to focus on proving the correct product shape and user interaction model.
Deterministic output schema
Defined a structured schema for obligations, risks, and tasks to ensure outputs can be operationalised rather than interpreted manually.
LLM-ready architecture
Designed with a clear upgrade path to integrate LLMs with structured JSON output, validation layers, and consistency enforcement.
Security Considerations
Planned architecture includes secure handling of sensitive regulatory data, with future integration of authentication, access control, and audit logging to meet compliance requirements in regulated environments.
Testing Strategy
Focused on product validation rather than full test coverage in MVP phase. Future iterations would include schema validation tests, LLM output consistency checks, and end-to-end workflow testing for compliance use cases.
Outcome
A working AI-native prototype demonstrating how unstructured regulatory text can be transformed into structured, actionable workflows. Validates the concept of moving from 'information tools' to 'decision systems' in compliance environments, with a clear path to production-ready architecture.
Interested in working together?
Let's discuss your next project.