Artificial intelligence in asset management: lessons for philanthropy

March 11, 2021

Reading Time: 5 min

The use of artificial intelligence in finance and asset management has by now made its case by producing returns, but the problems of opacity and hiring the right talent remain. What implications for philanthropy?

AI and finance: no longer “emerging” but no easier to grasp

If artificial intelligence (AI) is still in its infancy in philanthropy, with the Altruist League being one of the pioneers, then in the broader field of finance and, in particular, asset management, it seems to be nearing the point of maturity. This is in particular true for three areas: portfolio management, trading, and portfolio risk management.

We are already seeing in-depth assessments of the application of machine learning (ML) in the field produced by the likes of the CFA institute; financial institutions and consultancies are flooding the market with observations and best practices. From these management-friendly texts it would seem that the topic has somewhat lost its aura of obscurity and is becoming more approachable even to the non-expert.

In fact, the opposite is true. AI in finance is becoming ever more sophisticated, and ever more techniques are being employed. So far, the best hands-on text I’ve seen is Machine Learning in Finance: From Theory to Practice, by Dixon, Halperin and Bilokon. To someone with a solid background in mathematics and probability theory it takes a good month to go through. Others better not even try.

What are some of the key takeaways for the leader of an asset management organization, be it for-profit or a charitable foundation?

Proliferation of AI/ML techniques employed

Machine learning is a of course a broad concept that envelops many different techniques. A number of them are now finding their use in different aspects of asset management. Artificial Neural Networks (ANNs) are seeing use in forecasting, as are the LASSO techniques. The former are unsurprisingly seen as the most promising candidates for predicting stock returns.

Cluster Analysis is being employed to help with asset classification. Natural Language Processing (NLP) is exceptionally useful in trawling through company annual reports and news articles, including social media.

The Altruist League, for example, relies on it heavily on NLP for its model. Unlike more traditional textual analysis techniques, such as dictionary-based approaches that extract information only from individual words in the text, AI approaches can also interpret context and sentence structure.

Managers must, at the very minimum, familiarize themselves with the techniques being used in the field today and understand the broad capabilities of each one.

Scramble for talent

Applying quantitative models to finance is not new, and a background in science and mathematics has been prized in many trading and asset management circles for decades. However, the advent of artificial intelligence and machine learning raised the game to a whole new level.

We are witnessing a sub-specializations among data science experts in ways which we have already seen in software engineering. A specialist in reinforcement learning might not be the right person to train your ANN, and vice versa. Salaries may vary significantly depending on how specialized the sub-field is and how sought-after the talent.

Managers should appreciate the effect of this scarcity of key talent on their hiring practices, as well as the broader impact on the way the organization is structured and operates.

Opacity, complexity and risk

The black-box nature of many AI systems makes oversight and regulation (as well as compliance) difficult. This poses legal and security risks, as well as, naturally, risks for the bottom line. For example, ANNs that are trained to pick stocks with high expected returns might select illiquid, distressed stocks through incorrect inferences. Data quality and sufficiency can be other major sources of concern.

It is uncomfortable to bank your business on an algorithm that even your data scientists cannot understand. It is even more uncomfortable to entrust your retirement savings to such an organization. Unsurprisingly, company promo materials stretch themselves to emphasize human oversight of the entire process.

Robo-advising

Most of the developments in AI in finance are technology-driven, but some have a significant component of demand. Robo-advising in particular is popular among the millennial crowd, technology-savvy and unwilling to pay huge management fees to the traditional asset manager. Many new platforms have sprung over the past few years, and the growth of robo-AUM has been staggering, including during the pandemic – they are projected to pass $1 trillion in 2021.

Robo-advising can take several different shapes. The B2C model is the most prominent, certainly, but there are derivatives. At the Altruist League, we use a hybrid B2B model, where the AI advisor merely informs the human analyst who is in charge of taking the final decision on portfolio structure, financial allocations, and other matters.

Summary

With all their flaws, machine learning models in finance are here to stay. They can learn from large amounts of unstructured data and extract patterns invisible to humans. They excel at repetitive tasks and are designed to improve themselves by readjusting in accordance with the data.

However, managers, including foundation CEOs venturing into AI to improve their philanthropy, need to understand the implications: rather than being an asset management company that has an AI department, by embracing machine learning in our core business model we are becoming technology companies that do asset management.

Of course, not all use of AI must be focused on portfolio allocation, generating alpha or managing risks. There are advantages to be had from enhancing operational efficiency, improving content and product distribution or simply engaging better with clients. This might be a sensible place to start for organizations only beginning to formulate their artificial intelligence strategy.