Why allocators need a ‘continuous exploration’ mindset for AI adoption

Asset owners are witnessing a sea-change in data management and analytics as artificial intelligence opens doors to more efficient processing of investment information. But despite the promise, some may be wary of experimenting with the new technology due to the potentially negative consequences of failures on budget and culture.  

This tension was recently explored at the Top1000funds.com Fiduciary Investors Symposium. Jon Webster, senior managing director and chief operating officer at the C$777 billion ($554 billion) CPP Investments encouraged his allocator peers to first shift their mindset to one of “continuous exploration” when it comes to AI.  

He challenged the view that AI experimentation is inherently expensive to implement, at least for an organisation of the size of CPP Investments. It subscribes to the enterprise version of ChatGPT and uses DealCloud as its CRM software, whose combined fee is less than C$5 million a year against a C$2 billion annual operating budget and a C$10.5 billion investment budget.  

“I don’t think you should be afraid of the cost aspect of it. In fact, I’d lean very heavily into that,” Webster told the Fiduciary Investors Symposium at Oxford University.  

“To the experimentation point, it’s not that you should experiment. You literally have no choice but to experiment, because what the models can do tomorrow is just not the same as they were able to do yesterday.” 

To understand what AI can do for a fund, one needs to first understand what constraints it will remove, Webster said. He proposes that an organisation which is formed around investing is essentially “an information-processing supply chain” – it takes structured and unstructured data and turns them into investing ideas. 

The limitations, in turn, are “scarcity of expertise”, “high degrees of specialisation”, and “limited information processing capacity”, which is an intrinsic part of being a human.  

“Lots of those constraints are about what we prize deeply, as in ourselves, our own intelligence,” he said. 

“The thing we’re looking at inside CPP Investments is, how will our organisation get changed by this [technology]? How will work legitimately get redesigned? And are the constraints around which our organisational structure was formed, as valid as they were?” 

The C$26 billion ($18.5 billion) OPTrust has also been on a journey in the past six years to more comprehensively implement AI in its investment process. It started with using AI to determine risk on/off positions but now has a regime model of four macro quadrants upon which it makes asset rotation decisions, thanks to more accuracy in risk factor predictions enabled by large language models.  

“I don’t think we think about it from a perspective of, because this is a new technology, we have to chase it,” said Jacky Chen, managing director of completion portfolio strategies, total portfolio management at OPTrust.  

“It’s a very natural progression that as we think about our investment approach, we naturally gravitate ourselves towards that kind of decision making.” 

The most important principle of AI implementation at OPTrust is that a “human is always in the loop”, Chen said, but he acknowledged that the technology becoming a more prominent part of the investment process means portfolio managers also need to change their old behaviours.  

“As an example, when AI doesn’t know what to do when you ask it to build a risk indicator, it will immediately assume that risk equals to 0.5 times the fear index plus 0.5 times maybe the gold price. And they draw this beautiful chart,” Chen said. 

“[But] would you be able to make decisions on that? I think the traditional, more fundamental managers need to start understanding how to vet that and look into that [underlying model].” 

Conversely, quant portfolio managers need to be more market-driven, instead of treating market events as “noise” and focusing purely on their data sets as they traditionally have been.  

“The cycle is moving way faster now. If you don’t do that, you’re not able to very effectively manage your model,” Chen said. 

CPP Investments’ Webster said for asset owners, using AI is not about making faster but better investment decisions, “unless fast is your competitive edge”. 

“For many of us, it isn’t,” he said. 

“You’ve got to do something different with the technology, otherwise, it’s all a bit pointless.”