Trusted AI operating models
Strategy, risk, technology, capital, and people moving together as intelligence becomes infrastructure — at the scale of a systemically important bank.
The Philosophical Ledger / AI / Banking / Agents / Institutions
Intelligence is becoming infrastructure.
I lead AI and agentic platforms at the Middle East's largest bank. The Philosophical Ledger is where I think in public — a practitioner's account of how institutions move from AI experimentation to governed execution. Not a spectator's view of deployment. A builder's.
Read the essays, scan the situation room, follow the podcast map, and connect the signals that matter before they become consensus.
New Here? Three Doors In
Start where your question is.
Place your organisation
The Levels of an AI‑Pilled Organization
Six levels, four diagnostic questions. Find out where you actually stand — not where the press release says. 14 min.
Read the essay → Read today's front pageThe AI Situation Room
Verified AI news daily, placed on the five-layer stack — energy to agents — so you see where the pressure is building.
Open today's edition → See what builders believeThe Claims Ledger
Hours of AI podcasts compressed into scored claims — what builders, labs, capital and enterprise are converging on.
Open the ledger →Find Your Level
Where does your organisation actually stand on AI?
Six honest questions — four about your organisation, two about your agents. Your real level is your weakest answer; the asymmetry is the diagnostic.
Take the 6‑question diagnostic →Latest Thinking
I write down what is changing while it is still arguable.
Essays and market notes from two years building agentic systems inside a regulated bank — the operating-model changes serious institutions can no longer postpone.
Field Notes
The Amnesia Tax
Every firm pays an invisible tax: AI that captures the answer and forgets the judgment behind it, so nothing compounds. The escape is a governed learning loop — and why governance is the drivetrain, not the brake.
June 2026 / 16 min read Operating ModelsThe Levels, Revisited for Agents
The four questions diagnosed whether your org was AI-native. The agentic era forces two more — Trust and Memory — that decide whether your agents are more than a demo.
June 2026 / 5 min read Agentic AI StrategyThe Gap Is The Equation
80% of enterprises are building AI agents. 6% are getting results. The missing layer is architecture, governance, and institutional knowledge.
April 2026 / 8 min readAI Situation Room
My read on what is actually moving in AI — verified, placed on the stack, daily.
Track what changed, why it matters, and who should act across models, agents, regulation, banking, cloud, chips, government, and the GCC.
What I Believe
AI will expose weak institutions before it transforms strong ones.
The next advantage is not model access. It is judgment, trust, workflow ownership, and speed of learning.
Human judgment becomes more valuable when routine cognition becomes cheap.
Institutions will not be replaced by AI; they will be judged by how well they govern it.
The future belongs to people who can turn uncertainty into operating systems.
What I Build
The layers between an impressive demo and a system a regulator will sign off on.
Agentic control planes
Identity, authority, memory, observability, audit, escalation, and the right for systems to act.
Institutional knowledge systems
The semantic layer that turns scattered data, decisions, and memory into organizational intelligence.
Public intelligence products
Essays, market maps, radars, and executive briefs that help leaders see what is changing early.
What I'm Working On
Open-source systems for agents that need traceability, control, and trust.
The current work is focused on a simple problem: agents should earn trust through traces, evaluation, memory, and approval logic, not through polished answers alone.
The GitHub work follows that thesis. Ninja Harness evaluates the full execution trace so teams can see whether an agent should be trusted. Agent OS turns that into a runtime: profiles, skills, memory, traces, approvals, and a controlled improvement loop.
Books
A shelf for intelligence, judgment, systems, and human behavior.
A working library across AI, design, markets, history, money, statistics, decision-making, and the long arc of human institutions.
About
I build systems that let institutions trust AI enough to act.
I lead AI and agentic platforms at First Abu Dhabi Bank — the largest bank in the Middle East — where the work isn't demos but governed systems that have to survive real regulators, real risk, and real scale. Eighteen years across investment banking, fintech innovation, and enterprise AI — including eleven in Debt Capital Markets and fintech innovation at Bank of America — taught me how financial institutions actually decide: slowly when trust is missing, quickly when architecture and accountability are clear.
I also build in the open. Ninja Harness and Agent OS are the same argument made in code — that agents should earn trust through traces, evaluation, and approval logic, not polished answers.
I'm drawn to AI, philosophy, physics, and systems thinking because each asks one question: how do complex things become understandable enough to improve? Beyond the work, I'm a father and a permanent student.
Gartner Innovation in Financial Services judge (2024, 2025). Published researcher in Springer and IEEE-affiliated venues. Writing collected here as The Philosophical Ledger.
Subscribe
Get The Philosophical Ledger.
One brief, when there is something worth your time — on agentic AI, governance, and the operating model for systems institutions can actually trust. No filler between issues.