Blog
Policy-driven agents, document intelligence, and enterprise automation, from the team building it.
106 articles
The three ratios that size a commercial real estate loan, how each is built from the rent roll, operating statement and appraisal, the supervisory LTV limits, and what ongoing monitoring examiners expect.
What financial statement spreading is, how analysts map borrower financials into a credit template, why layout changes and non-standard statements break template tools, and what examiners expect of the analysis.
What a lender checks on every construction draw: G702 and G703, invoices, lien waivers, inspection, title update and change orders, and what examiners expect.
How loan covenant testing works: what the credit agreement defines, how each period is tested, where manual tracking breaks, and what examiners look for.
Build the exam file that maps three NYDFS AI letters onto 23 NYCRR Part 500: risk assessment, access, vendor contracts and incident-response evidence.
McKinsey's 2025 State of AI: 88% use AI regularly, 62% experiment with agents, yet only 39% report any EBIT impact. Why the savings stay hidden.
A deployer under the EU AI Act uses an AI system under its authority. White-labeling, substantial modification or a new purpose can make it the provider.
NIST AI 600-1 names twelve generative AI risks. File an owner, evidence and residual-risk acceptance for each before a tool-calling agent goes live.
Agentic model risk is the exposure left after OCC Bulletin 2026-13 placed generative and agentic AI out of model-risk scope. What banks still need to govern.
Treasury released the FS AI RMF in February 2026 with 230 control objectives. Start with inventory, owners, data, oversight, vendor access and incidents.
EU AI Act Annex III high-risk duties now apply from 2 December 2027. 2 August 2026 still brings Article 50 transparency and enforcement.
Third-party AI risk is the exposure a bank inherits when a vendor ships an agent, or turns on agent features inside a release already under contract.