By Keith Viverito, Managing Director, EMEA of Clearwater Analytics

For decades, winning an institutional investment mandate involved the all too familiar examination of performance, investment philosophy, risk controls and of course fees. But if this wasn’t enough, fund managers now have to brace themselves for the inevitable question of “what, exactly, are you doing with artificial intelligence (AI)?”

Like it or not, AI is rapidly becoming part of the institutional due diligence process and could ultimately play an increasingly important role in determining which managers win billions of pounds of pension, insurance and sovereign wealth capital. Our recent research among 178 senior executives at asset managers across Europe, the US and Asia found that 90 per cent expect allocator scrutiny of AI adoption to increase over the next three years. More than a third expect it to increase dramatically.

The reality of today’s investment world is that a large pension scheme deciding where to place several hundred million pounds is not merely buying investment performance. It is buying confidence that the fund manager can continue producing it. Increasingly, that means understanding whether technology is genuinely improving the investment process rather than simply decorating the corporate presentation. The precise questions will inevitably differ. A pension scheme, insurer and sovereign wealth fund operate under different constraints and pursue different objectives. But the direction of travel is similar: managers will increasingly have to demonstrate what AI is actually contributing to the investment process. This creates an interesting problem.

For several years now, the asset management industry has been under pressure to demonstrate that it has an AI strategy.  Much of that pressure is now coming from within asset management groups themselves as firms look for greater efficiency, lower operating costs and ways to make better use of increasingly large volumes of data. Our research also found that 95 per cent of firms increased AI spending over the past 12 months, while 85 per cent plan to increase their budgets by at least 50 per cent in the year ahead. Allocators simply add another source of pressure because they increasingly want evidence that these investments are producing something tangible.

The danger now is that managers confuse having one with having a good one. Buying the latest technology is relatively easy. Making it useful is considerably harder. The asset managers reporting stronger results from AI are not necessarily those spending the most or those that started earliest. The report finds benefits in deeper analysis, sharper risk management and faster reporting, while identifying the underlying data as the crucial foundation.

This is where allocator questioning is likely to become uncomfortable. “Do you use AI?” is an easy question. However, “show me where AI improved an investment decision” is much harder. So are questions about what data fed the model, how an output was checked, who remains accountable when it is wrong and whether AI is genuinely improving risk adjusted returns rather than simply allowing the same work to be completed faster.

There may soon be another twist as AI could decide which managers are worth asking in the first place. Institutional searches can begin with long lists of potential managers and extensive RFP responses. It is therefore not hard to imagine pension schemes, investment consultants and other asset owners using AI to help interrogate those submissions and narrow the field. Who knows, in the future, pitches may need to convince an algo before convincing an investment committee.

There is, of course, an important balance here. Institutional investment remains a human business. Asset owners want access to portfolio managers and confidence in the people ultimately responsible for their capital. AI should not remove that accountability. But if it can make due diligence faster and help allocators identify the questions that deserve greater human scrutiny, it is likely to become increasingly difficult to ignore.

That distinction will become increasingly important as AI moves closer to the investment decision itself. Once machines are helping analysts interrogate larger datasets, identify patterns and test investment ideas, AI becomes more of a part of the investment process as opposed to principally an IT expenditure. And anything influencing investment decisions eventually becomes part of investment governance.

The irony is that asset managers spend their lives interrogating companies about capital allocation. Did management invest wisely? What return did it generate? Where is the evidence? Allocators are about to ask fund managers much the same questions about AI.

That is, in a way, all very healthy. The winners will not necessarily be the firms with the biggest AI budgets, the most impressive demonstrations or the greatest number of models. Instead, they will be those capable of showing where the technology makes humans better at investing, controlling risk and serving clients.  And if AI itself starts helping allocators draw up their shortlists, the consequences of failing to explain that clearly could become even more immediate. Institutional mandates have always come down to a beauty parade, and AI has just added a new round to it. “We are exploring AI” is not going to be enough to get through.