The traditional portfolio manager, meticulously sifting through a hefty stack of research and diligently inputting annual reports into a valuation model, seems to be a fading image from the past. For banks and asset managers to stand out, they must wholeheartedly adopt artificial intelligence, posits Gerben de Zwart of APG, one of Europe’s largest asset managers, overseeing approximately 541 billion euros for Dutch pensioners.
«Despite Dutch entities lacking the budgets and influence that global giants like Blackrock wield, even with a handful of specialists, you can still outpace your rivals,» opined APG›s Managing Director of Quantitative Strategies at the sixteenth Fondsevent, held on Monday in Bussum, the Netherlands. During the morning segment, De Zwart shared insights alongside computer scientists Max Welling and Cynthia Liem.
Three years prior, the pension investor embarked on an ambitious goal to rejuvenate its investment department through a significant digitisation initiative, setting aside a budget of 100 million euros. «If we don’t embrace extensive digitalisation, our products will turn generic,» De Zwart reflected on their rationale back then. «Our aim was to achieve digital maturity by 2025, allowing us to operate more efficiently and invest more astutely.»
Investment strategy as pivotal
He reminisced about the archetypical portfolio manager 1.0 – an individual engrossed in gleaning and analysing data from company annual reports and executive meetings. Such professionals would often find scant time for what’s truly pivotal: devising the investment strategy.
A few years ago, APG made a strategic decision to transition partly from this traditional portfolio manager model to automated systems. This shift intended to free up time for high-salaried portfolio managers to refocus on investments and pose the right queries. «It’s not about where or how to procure quality data, but discerning the quintessential questions and data sources,» De Zwart elaborated.
On the topic of fully automating investments, De Zwart commented, «We don’t envision machines completely replacing investors at APG. We champion a blend of human intuition with machine efficiency. After all, humans are indispensable for guiding machines and infusing them with fresh perspectives. And for a long-term investor like APG, when you’re acquiring significant stakes in companies, the decision isn’t solely based on AI; it’s also about the people involved.»
Automated scan of 10,000 annual reports
Presently, APG utilises AI predominantly for sustainable equity investments. They employ an automated system to scan annual reports of approximately 10,000 publicly listed firms, relying on a predefined taxonomy. «Our approach is intricate. We align our objectives with each of the United Nations› sustainable development goals, detailing what complements or detracts from them,» explained De Zwart. «The beauty of this taxonomy is its scalability. Each year, we refine and augment it.»
Almost two dozen individuals are engrossed in fine-tuning this model at the pension investor’s end, although a substantial portion of the heavy lifting is shouldered by machines. De Zwart shared, «Computers disqualify about 80% of companies. Our in-house analysts then delve deeper into the remaining 2,000. Ultimately, it’s a human decision on whether to invest or not. The system merely streamlines the selection process.»
He perceives AI as a tool that provides an edge in information gathering. «This isn’t limited to merely dissecting annual reports; it extends to assimilating paid data sources, including credit card transactions and retail data, and even leveraging translation tools,» De Zwart added. «Understanding how data about one company might influence another is invaluable.»
Caution
However, a word of caution about APG›s AI-centric methods came from computer scientist Cynthia Liem of TU Delft. «Don’t be too hasty in delegating your processes. Always be cognisant of the nuances. A machine might rashly act on a minor discrepancy when human discernment is essential. Reflect meticulously on what you manage in-house versus what you delegate.»
Regarding using AI for translating global annual reports, Liem sounded another alarm for De Zwart. «Be wary of linguistic subtleties. A direct translation might overlook cultural or idiomatic nuances in, say, a Chinese annual report. An adept analyst familiar with the language might unearth hidden gems. Otherwise, you might inadvertently misplace that proverbial needle in the haystack.»
Related articles on Investment Officer: