Blog

AI agents, RAG, and integrations — notes from production.

First-Party Fraud in Lending: Detecting Intent Not to Repay

First-party fraud in lending is a real applicant who never intended to repay. Identity checks miss it entirely - the tell is behaviour, not identity.

Fraud Scoring for SaaS Billing: Trial Abuse and Card Testing

SaaS billing fraud differs from ecommerce: trial abuse, card testing via low-value subscriptions, promo abuse. Same scoring pattern, new features.

Card-Issuer Fraud Scoring: Authorization vs Post-Transaction Review

Card-issuer fraud scoring compared to merchant-side scoring: authorization-time decisions under real time pressure, plus a slower post-transaction review layer.

Fraud Detection for Cross-Border Remittance

Remittance fraud detection is less about card fraud and more about identity, corridor risk, and structuring. The sender-recipient pair is the entity to score.

Fraud and Chargebacks in Online Gaming and Gambling

Gaming and gambling fraud differs from ecommerce: bonus abuse, multi-accounting, deposit-play-withdraw laundering. Same scoring approach, new signals.

Marketplace Payment Fraud: Protecting Both Sides of the Transaction

Marketplace payment fraud is two-sided: buyer fraud with stolen cards, and seller fraud or payout collusion. Both need the same architecture.

Fraud Scoring for Buy-Now-Pay-Later: Underwriting at Checkout Speed

BNPL fraud detection is the hardest of both worlds: a fraud and credit decision made at checkout speed, with adverse-action rules on the credit side.

Fraud Detection for Crypto Exchanges: On-Chain and Off-Chain Signals

Crypto exchange fraud detection combines off-chain signals - device, KYC, funding, behaviour - with on-chain analytics on deposit and withdrawal addresses.

Fraud Scoring for a Neobank: The Onboarding and First-Transaction Window

Neobank fraud detection is hardest in the first days of an account - no behavioural history, so scoring leans on onboarding signals, device, and funding source.

Vendor Lock-In in AI: Where It Hides and How to Limit It

AI vendor lock-in hides in prompts and flows in a vendor console, a fine-tuned model you can't export, conversation logs, and embeddings tied to one model.