Risk Assessment Rule Management System

challenge

Challenge 1

brief

The team built a rule‑driven risk assessment system that converts SME survey responses into structured, validated heuristics. Using LLMs, fuzzy matching, and human‑in‑the‑loop review, they generate, deduplicate, and govern high‑quality risk and mitigation rules that can be applied consistently across risk registers.

Please be aware that this content was generated follwing an automated review so may not be perfectly accurate; refer to the original challenge brief and team files for authoritative information

key outcomes

Improved consistency and clarity of risk and mitigation definitions; reduced duplication and ambiguity in heuristics; stronger guardrails against low‑quality or hallucinated AI outputs; faster standardisation of SME knowledge into reusable rules.

important files

• deduplicate_rules.ipynb: Notebook for identifying, deduplicating, and merging similar heuristic rules using fuzzy matching.
• generated_rules.json: Structured JSON output of validated risk and mitigation rules.
• Categorisation of Risk Heuristic Survey Questions.docx: Defines categories used to classify SME heuristic survey responses.

details

team: Risk Assessment Rule Management System members: tbc topics: solution-centre, hack26, challenge1, python, jupyter, fuzzywuzzy, json, large-language-models, risk-management, heuristics, rule-engine, data-quality, governance, human-in-the-loop technologies: python, jupyter, fuzzywuzzy, json, large-language-models