Governed Autonomy

How Private Banks Can Industrialise AI Within Regulatory Boundaries

Published March 2026

The institutional altitude. A four-tier deployment architecture—machine learning controls, large language model workflows, ontology-bounded autonomy, and an institutional governance wrapper—mapped against the composite regulatory landscape facing Singapore private banks. The periodic client review serves as the exemplar productised journey.

Addendum suite (in progress): organisational structure and operating culture; reference architecture (open source); reference implementation (open source); enterprise value model; evaluation and assurance framework.

Governed Autonomy II

Regulating Autonomous Economic Agents Operating within Tokenised Financial Infrastructure

In progress

The agent–infrastructure altitude. Introduces lex machina—extending De Filippi's lex cryptographia—and a graduated taxonomy of agent deployments from Delegated to Supervised, Bonded, and Sovereign, each with distinct governance requirements. The autonomous robotaxi serves as the exemplar. Where Governed Autonomy takes the regulatory framework as given, Governed Autonomy II interrogates its adequacy where machine agency, code-as-law, and human accountability intersect at the infrastructure layer.

Addendum suite: Forthcoming.

Governed Autonomy III

The Regulatory System as a Control Problem

Planned

The systemic altitude. Treats regulatory institutional design itself as the unit of analysis, drawing on Leveson's STAMP as analytical engine and Ostrom's design principles as evaluation criteria for polycentric governance. Core questions: where the regulatory perimeter sits when governance is computationally embedded; how jurisdictional architecture coheres across MAS, the Federal Reserve, the PRA, the EU supervisory framework, and the HKMA; what systemic risks emerge from agent–institution–infrastructure interactions; and whether existing instruments remain adequate when rules, implementation, and verification share the same execution environment.

Addendum suite: Forthcoming.

AI Governance Readiness Assessment

Live

A diagnostic against Singapore's regulatory framework for AI in financial services, mapped specifically to private banks. 193 requirements across 7 domains, mapped to 13 source instruments spanning MAS statutory and supervisory expectations, IMDA frameworks, the ISO 42001 assurance standard, and PDPC guidance. Available as a 25-minute screening or a 90-minute full assessment.

AI Governance Economics Engine

In progress

The operational companion to Governed Autonomy. The tool computes the true all-in cost of an AI workflow—inference, human, audit, and governance—under the institution's specific operating assumptions; projects how that cost scales across a workflow portfolio and where amortisation occurs; compares cost under the three archetypes (Compliance First, Workflow Augmented, Governed Autonomy) so the unit economics thesis becomes calculable rather than rhetorical; and produces a board-ready governance gate pack that links cost projections to risk materiality and MAS-aligned controls.