As the very real threat from artificial intelligence (AI) to global security makes its way to the top of policymakers’ agendas, so new research shows asset owners may be applying the technology without accompanying oversight.
The Marsh Global Asset Owner Barometer 2026 finds 60 per cent of 74 asset owners with more than USD2 billion in assets under management identify AI tools and/or use cases as a capital-expenditure priority over the next 12 months. However, just 13 per cent have enterprise-wide AI controls and defined investment use cases in place.
Niall O’Sullivan Global Chief Investment Officer, Investments and Retirement at Marsh, says: “Asset owners are directing capital toward technology faster than they are building governance around its use across their organisation. Investment is accelerating across data, analytics, and operating systems, although mature systems and clearly defined investment applications remain uncommon. The gap between expenditure and governance reinforces the broader challenge of converting technological potential into repeatable investment capability.”
The study finds that asset owners’ deployment of AI is concentrated in information-intensive activities. Data management or knowledge retrieval leads at 47 per cent followed by workflow automation at 42 per cent.
Next is risk monitoring or scenario analysis at 41 per cent and manager research or due diligence at 39 per cent.
Direct investment portfolio applications are less common: 31 per cent use AI for portfolio construction or optimisation, and 14 per cent for trading, rebalancing, or execution support.
O’Sullivan notes: “The focus of organisations’ capital spending indicates where organisations expect capabilities to develop.”
Technology tools or applications are the leading priority at 58 per cent, ahead of data management and quality at 46 per cent, and portfolio analytics or scenario tools at 44 per cent.
While governance may lag asset owners’ implementation of technology, the resera4sh does show some effort to put risk controls and safeguards in place.
Cybersecurity controls (43 per cent) and human review of AI-supported outputs (43 per cent) are the most-cited AI-risk management approaches, followed by internal AI-use policies at 41 per cent; staff training at 39 per cent; and privacy controls at 36 per cent.
The study shows asset owners are examining the potential impacts of AI and technology disruption on investments and returns – only 7 per cent are not assessing AI exposure across their portfolios.
O’Sullivan says: “The maturity of AI governance across organisations lags the prominence of the topic as a strategic priority. While some organisations have moved beyond initial pilots to define specific investment use cases, most initiatives are yet to reach widespread adoption.”
The Marsh study coincides with concerns from major tech developers, including OpenAI and Anthropic, that are urging lawmakers to regulate AI, with Microsoft’s AI boss the latest to say rules are needed to prevent a “silicon species” from destroying humanity.
