Strategic Intelligence & INSIGHTS

A board that asks these twenty-five questions and records honest answers will produce, in a single sitting, the most accurate picture of its institution’s real AI control posture it has ever held. 

The pattern of answers — not any single answer — is the finding:

· Identify the critical cells. Five cells — one per ecosystem — typically carry the highest fiduciary consequence. Resolve those first.

· Assign an owner and a date to every "assured only" answer. An open control gap without an owner is an unmanaged risk.

· Make this a standing review. AI control is a fiduciary matter that sits above the technology-risk layer. The twenty-five questions are a quarterly instrument, not a one-time exercise.

The institutions that safeguard, administer, and report on the world's invested capital are no longer back-office utilities. Custodians, fund administrators, transfer agents, and securities services providers have become the operating layer the entire investment ecosystem depends on — and that layer is being rebuilt, in real time, on artificial intelligence.

Production AI is already moving into reconciliation, corporate actions, NAV production, tax, and client reporting. The defining question of 2026–2027 is no longer whether asset servicers will deploy it. It is whether they can demonstrate control over it — show, on a regulator's or auditor's timeline, exactly what an AI system did, on what basis, and on whose infrastructure.

This paper maps that gap. It analyzes the global accounting function stack one function at a time — trade capture through financial reporting — traversed across mutual funds, ETFs, hedge funds, private markets, retirement, and insurance general accounts, with the regulatory divergence between the US, UK, EU, and APAC flagged where it changes the operating model. Every named assertion is sourced.

The view it advances is consistent across every category, function, and jurisdiction: in our assessment, AI control — not AI adoption — is emerging as the binding operational constraint on the industry. The infrastructure is being built at remarkable speed. The control discipline is not yet keeping pace.

An Institutional AI white paper · May 2026 · 43 pages · Reference edition

Institutional asset owners enter 2026 confronting a fundamentally different operating environment. Global pension assets reached a record USD 68.3 trillion. Sovereign wealth funds posted one of their strongest performance years on record. U.S. higher education endowments returned 10.9 percent. Yet the structural risks beneath those numbers have intensified.

Our annual research synthesizes findings from 25+ industry studies — BlackRock, McKinsey, Mercer, WTW Thinking Ahead Institute, Invesco, UBS, NACUBO-Commonfund, Natixis, KPMG, and others — into a category-by-category analysis across pension funds, sovereign wealth funds, insurance companies, family offices, and endowments and foundations.

Three structural forces define the 2026 — 2027 horizon: the convergence of public and private markets, the return of active management, and the AI governance/control gap that has become the binding constraint on institutional readiness.

Intended for boards, investment committees, CEOs, COOs, CIOs, and business unit leaders at the world's leading asset owners.

[See the full report]

27 pages • May 2026

   Institutions face an existential paradox: they rapidly adopt artificial intelligence across operations and strategy while operating under a dangerous illusion that using AI infrastructure means controlling it. In reality, 78% of enterprises run mission-critical AI workloads on third-party platforms they cannot audit, in jurisdictions they do not govern, powered by energy sources they do not control.

The concentration of AI production has reached unprecedented levels. 92% of advanced AI chips are fabricated by a single company in Taiwan. 70% of global AI compute capacity is controlled by five providers. By 2030, AI and data-center operations will consume 945 terawatt-hours annually—more electricity than Germany. Intelligence sovereignty and energy sovereignty are now inseparable.

This paper introduces The Five-Pillar Control Framework—jurisdictional, logical, technical, operational, and contractual control—demonstrating how institutions transform regulatory compliance into measurable, continuous sovereignty. At the center of the analysis is the emergence of AI factories: massive physical installations consuming 100-500 megawatts each, where energy, compute, data, and models converge to produce artificial intelligence at scale. These facilities are not metaphorical. They are the refineries of the 21st century, and the institutions that control them will define geopolitics and institutional power for the next half-century.

Drawing on data from the IEA, OECD AI Policy Observatory, Uptime Institute, and binding regulatory frameworks (GDPR, NIS2, HIPAA, EU Data Act, EBA Guidelines), the paper provides a maturity assessment framework and a decision architecture: Should institutions build sovereign AI infrastructure, rent with enhanced governance, or compose a hybrid model? The assessment integrates regulatory requirements, strategic AI dependence, risk tolerance, and financial capacity to determine the appropriate path—with specific thresholds ranging from standard cloud acceptable (score 0-40) to sovereign infrastructure mandatory (score 120-160).

The conclusion is binary: AI has transformed governance from a matter of oversight into a matter of design. Institutions that embed control—technically, operationally, architecturally—will command intelligence. Those that do not will be commanded by it. Ownership is optional. Control is non-negotiable.

[See the full report]

85 pages • November 2025