Insights
Essays on banking AI that pays its way.
Frameworks and analysis for the executives accountable for AI spend — written to be argued with in a boardroom, not skimmed.
The Vendor Concentration Heatmap: You Are Not Diversified
Your AI estate has three vendors and may have one dependency. Two supervisors have now independently named substantially the same short list of critical providers — and the six layers where concentration actually sits are not the layer your supplier count can see.
The AI Unit-Economics Ledger: Why Cheaper AI Costs More
Model prices collapsed and bank AI bills did not follow, because a bank does not buy tokens — it buys completed work, and the four recurring cost lines that decide what that work costs make this a forecasting problem rather than a cost one.
The Payments Defense Grid: The Control Worked. The Money Still Left.
A payment control is worth what it can see, not how fast it fires. Money leaves an account wrongly in only two ways, and no amount of investment in impostor controls buys any coverage of the second — two axes, six cells, and the half of a payment defence most banks have never built.
The Grounded-Knowledge Architecture: Why a Cited Answer Can Still Be Wrong
Citation presence, claim support and current applicability are three different properties of an answer. Five failures hide behind one word, each with a different owner — and six capabilities, proposed not ratified, that decide which one just happened.
The Agent Attack-Surface Map: Where an AI Agent Actually Gets Attacked Inside a Bank
Five surfaces and one control layer. The exposure is not the model, the surface with no owner is the tool's own manual, and the cheapest control on the map is the permission you have not granted yet.
The Adverse-Action Explainability Ladder: What a Bank Owes an Applicant It Rejects
A sound model, a failed operating model, a servicing bill of over 89 million dollars. R3 is the floor, R4 is the trap, and Brussels is the last clock.
The Deposit Loyalty Gate: Which Deposits Agentic AI Takes First — and the Gate That Holds Them
Reachable is not the same as gone. The Deposit Loyalty Gate: two axes, an authorization gate, and the Vulnerability Quadrant that names which deposits leave first.
The GenAI Assurance Grid: How to Govern AI You Cannot Open
SR 26-2 puts generative AI outside model-risk scope, not outside governance. The GenAI Assurance Grid: five rings that assure the system you cannot open.
The AI-Augmented Bank: A Five-Pillar Enterprise AI Operating Model
A more digital bank is not a smarter bank. Read the AI-augmented bank as five pillars in dependency order, then fix only the pillar that is binding.
The AI-Ready Bank: A 5-Layer Data-Readiness Map
Your data isn't AI-ready — and no model fixes that. Read the five layers that decide whether AI pays back, then remediate only the one that binds.
Stop Funding AI Pilots — Run the Portfolio Test
Banks fund AI as scattered pilots that never ship. The AI Use-Case Portfolio scores every use case on value × feasibility — then funds, gates, or kills it.
One Human, Thirty Agents — the Next Bank Operating Model
One human, thirty agents — a ratio you earn, not one you pick. The Agent Army Operating Model: four pillars, five forces, and the Span Audit.
The Build-Buy-Partner Grid — Stop Building Your Own Banking AI
Most banks are building AI they should have bought. The Build-Buy-Partner Grid — two axes, four verdicts, and the Sourcing Decision Loop for banking leaders.
The Operating-Model Multiplier — Why Two Banks Buy the Same AI and Only One Gets the Return
Two banks buy the same AI and one earns 5%, the other 50%. The difference is the operating model. The 3 levers, the 70/20/10 Law, and the Multiplier Audit — for banking leaders.
The Visibility Trap — Why Banking AI Is Deployed Where the Money Isn't
Most banks spend their AI budget where the win is easiest to see, not where it lands on the P&L. The Visibility Trap, the 3 Value Zones, and the Capital Reallocation Loop — for banking leaders.
The Velocity of Control — Why Banking AI Governance Makes You Faster
Manual governance is the bottleneck, not governance itself. The 4 Stone Guardians — Gate, Guardrail, Ledger, Owner — let banks ship AI fast and stay defensible.
How Far Should a Bank Let AI Act on Its Own? The Agent Maturity Ladder
Autonomy isn't a switch — it's a ladder. Four rungs, two rails, and the Explainable·Overridable·Accountable gate that decides how far a bank should let AI act on its own.
Why 94% of Banks See No Value From AI — And the Boardroom Equation That Fixes It
Banks spend billions on AI but 94% see no measurable value (McKinsey 2025). The Boardroom Equation makes AI ROI auditable in three pillars.