RESOURCES

Insight and
domain guides

Current thinking and working guides on insurance RegTech, credit risk, defense AI and supply chain AI agents.

  INSURANCE REGTECH

Insurance regulation and AI

Insurance RegTech

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.

Insurance RegTech

ICS · RBC and IFRS 9 data quality rulesets — checking what your regulatory calculations assume

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.

Insurance RegTech

Why actuarial work needs explainable AI

Why auditors will not accept black-box AI, and how an XAI-backed audit trail resolves it.

  CREDIT RISK

Credit risk and high-impact AI

Credit risk

PD is not one number — what the end of a grace period does to the risk curve

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.

Credit risk

Eight things a high-impact AI underwriting system needs — a working risk-management matrix for Korea's AI Framework Act

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.

Credit risk

Thin-file lending — what you lose by treating a missing value as a deduction

How the handling of missing values shapes the outcome for small businesses and sole proprietors with thin files, and a confidence-adjustment approach instead.

  DEFENSE AI

Defense AI and specification management

Defense AI

Parsing HWP — why foreign solutions can't get into Korean defense

The reality that 90% of Korean military documents are HWP, and the technical approach to parsing them accurately.

Defense AI

Building an LLM on a closed network — running AI air-gapped

The architecture for running an sLLM on an on-premises server with no internet, and how it has actually been implemented.

Defense AI

Automating gap analysis — turning 2–4 weeks of specification review into tens of minutes

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.

  SUPPLY CHAIN AI

Supply chain AI agents

Supply chain AI

Cross-verifying import customs documents — what the agent actually looks at

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.

Supply chain AI

Adding agents without changing your ERP — connector design principles

A working reference on integrating without touching the database directly — registering your APIs as tools, and keeping source data inside the company.

Supply chain AI

AI judges, people decide — designing the human approval gate

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.

We'll tell you when something new is published

Leave your email and you'll be first to get new material on insurance RegTech, credit risk, defense AI and supply chain AI agents.