A checklist for automating assumption setting and validation — 10 things to check
A working guide to finding the manual stretches of assumption mapping, validation and approval that survive a CF engine rollout — and deciding what to automate first.
Current thinking and working guides on insurance RegTech, credit risk, defense AI and supply chain AI agents.
A working guide to finding the manual stretches of assumption mapping, validation and approval that survive a CF engine rollout — and deciding what to automate first.
The source-data consistency problems that recur across Korean ICS · RBC and international IFRS 9 projects, and a ruleset-based approach to automating validation and cleansing.
Why auditors will not accept black-box AI, and how an XAI-backed audit trail resolves it.
The time structure a point-in-time PD misses, explained in working language through the term structure of a survival model. Conceptual throughout; no institution's figures are used.
How to satisfy risk management, explanation, human oversight and prior notice inside the underwriting screen itself, item by item — alongside the right to explanation under the Credit Information Act.
How the handling of missing values shapes the outcome for small businesses and sole proprietors with thin files, and a confidence-adjustment approach instead.
The reality that 90% of Korean military documents are HWP, and the technical approach to parsing them accurately.
The architecture for running an sLLM on an on-premises server with no internet, and how it has actually been implemented.
How an old and a new revision are compared automatically, how change types are classified and risk-graded, and why citing the governing clause is not optional.
How invoice, packing list, B/L and C/O are reconciled against the order and contract terms, and how each mismatch is passed to a person for approval with its reasoning attached.
A working reference on integrating without touching the database directly — registering your APIs as tools, and keeping source data inside the company.
How to widen autonomy step by step — recommendation, then execution after approval, then conditional automation — while keeping every binding act behind a person's approval.
Leave your email and you'll be first to get new material on insurance RegTech, credit risk, defense AI and supply chain AI agents.