Asset allocators struggle to assess AI risk because exposure already sits across portfolios in infrastructure, private assets, and manager behavior, often without clear visibility or measurement.
Paul O’Brien, a Member of the Wyoming Retirement System's Investment Committee, frames AI as a governance and oversight challenge first, not a stock-picking exercise, with impacts that cut across every asset class.
He points to using AI itself for exposure mapping, manager verification, and private market transparency while favoring foundational infrastructure and preparing for sustained volatility.
For institutional asset allocators, the loudest questions about AI often miss the one that matters most. While debate swirls around bubbles, breakthroughs, and headline winners, the AI exposure—and associated risk—already sitting inside portfolios remains hard to see, harder to measure, and rarely mapped with any precision.
Paul O'Brien, a Trustee and Member of the Investment Committee at the Wyoming Retirement System, brings decades of experience across economics, central banking, and institutional investing. His career includes senior roles at the Abu Dhabi Investment Authority and Morgan Stanley Investment Management, as well as time at the Federal Reserve Board, where he regularly briefed Governors on financial markets. Today, his perspective is shaped less by market cycles than by fiduciary oversight and long-term stewardship.
"The single biggest issue facing asset allocators today is how they should play the AI theme. But before you can play it, you need to know what your exposure is. The big challenge being missed is how to actually measure that AI exposure, because it goes way beyond Nvidia and the Magnificent 7. It affects almost every asset class and every piece of your portfolio," says O'Brien.
He advocates a classic "picks and shovels" strategy, focusing on the foundational infrastructure required to power AI. The approach means identifying exposure in real-world assets like the data centers springing up in states like Pennsylvania and the massive upgrades needed for power transmission. Private investment in digital infrastructure is now viewed as a utility-like asset class.
Power plays: "A new nuclear plant takes six to eight years to come online. The AI bubble will have passed before then," explains O'Brien. "That's why I see an investment like that as a short-term trade, not a structural holding. The closer you get to the copper wires and the transmission grid, the more it becomes a structural holding. Because whatever technology generates the electrons, you are the one carrying them."
For O’Brien, this search for exposure extends to the structure of the investment industry itself. He says that as a general-purpose technology, AI's primary impact won't just be automating tasks but reconfiguring how the industry is organized. The layers of intermediaries, he notes, create a "huge amount of drag between our assets and our investments from asset managers, consultants, and index providers," a friction AI is positioned to reduce. "Much of that will be disrupted and transformed by AI."
An end to opacity: O'Brien believes the disruption will be especially significant for private markets. That opacity has long been a defining feature of the private markets, which now face new pressures from limited partners demanding more information. He suggests AI is the force that will break the model open. "AI's biggest impact will be on private assets, because they won't be private anymore. These markets have thrived on opacity and a lack of transparency. But AI frees up access to information and analysis, giving limited partners true visibility into their portfolios down to the asset level. AI will disrupt, if not destroy, the illiquidity and opacity gap that separates private and public markets, giving LPs far more information and power."
AI audit: For allocators, using AI as a proactive oversight tool can be an important step in operationalizing governance. The approach allows for direct oversight, even as firms navigate the internal challenges of AI adoption. Effective oversight requires robust data governance to create a secure environment where AI can be used for accountability. "An interesting area right now is manager selection," notes O'Brien. "An individual investment officer might review a few dozen managers a year; AI can screen hundreds or thousands. More importantly, it allows you to verify if a manager is doing what they said they were going to do. You can use AI to analyze every trade and see if they follow the style that the manager was supposed to bring. Manager selection is a huge opportunity to cut costs and to improve outcomes."
So how should leaders navigate the road ahead? O’Brien's framework is grounded in a disciplined, three-part approach: study financial history, follow the technology's progress, and prepare for volatility. He notes the AI boom could even have a positive effect on pension liabilities, much like the 1990s internet wave, which saw higher productivity depress inflation while high capital demand pushed up interest rates—a net positive for pension balance sheets.
Back to the future: "I would study past investment bubbles, from the internet to railroads in the 19th century and electricity at the turn of the 20th. We have done this before; this is not new. A study of economic and financial history is really important here."
The danger, O’Brien argues, isn’t misjudging the timing of an AI boom or bust. It’s mistaking the shape of what comes next. AI’s impact on markets won’t unfold in clean cycles or orderly phases, and allocators who expect a smooth trajectory risk losing discipline just when it matters most. "My issue with the bubble analogy is that it implies something smooth, that grows in a stable way. That is not how the next few years are going to play out. There will be enormous volatility, and if you’re not prepared for it, you’re going to get thrown off your strategy," he concludes.
