AI for Money Managers: Avoid the Black Box – And Do This Instead

Humans have long worried that they would eventually create a technology they couldn’t control – and, at least to some extent, those worries have materialized. That’s true in the investment sector as well. We’ve heard stories about how artificial intelligence is able to “pick winners” and make overnight fortunes for investors  – but even top scientists often have no idea how AI is doing those things.

This “black box” dilemma has significant implications on many levels – including predictability and enhancing risk management, knowing when to invest and when to divest, one of the most important issues. And that predictability issue is especially acute when it comes to financial management – especially institutional investing, which could have a major impact on entire markets, as well as the savings and assets of hundreds of millions of people. If institutional investors don’t fully understand how their AI solutions work, how can they (and their clients) trust it to make investment decisions?

On the other hand, there’s no doubt that AI could be used to  enhance profits – and in fact, many institutional investors are already using it to find better ways to invest their organization’s assets. Many investors concentrate on specific assets, using AI to time purchases and sales – to great success.

The challenges slowing the adoption of AI

In theory, what works on a “micro” level could work even better on a “macro” level – where AI is applied to a wide variety of investments and makes recommendations based on massive amounts of data, using machine learning and other AI techniques to compare current market and world conditions to previous data, and determine which assets are likely to rise or fall in price based on that analysis. The opportunities afforded by AI are truly significant – but can we trust black box AI to produce the right results?

For many institutional investors, the answer is likely to be no – that the potential benefits of AI just aren’t worth the risk associated with a process they aren’t able to understand, much less explain to their boards and clients. As long as AI is making money for an investor, of course, no one will ask for that explanation – but if things go south, institutional investors will have to produce clear reasons as to why they made specific decisions. For many institutions, saying “the computer told me to” is unlikely to be a satisfactory answer.

Embracing transparency and a platform approach

But the alternative – avoiding AI – isn’t a viable path either. Other institutions that are less cautious, and do utilize AI, will likely do better on a wide range of assets – and then boards will be asking investors why they are leaving potential profits on the table, for their rivals to scoop up.

But there is a way out of this dilemma. Instead of utilizing AI systems that they cannot explain – black box AI systems – they could utilize AI platforms that use transparent techniques, explaining how they  arrive at their conclusions. AI systems do deep-dive analysis on huge reams of data, employing sophisticated algorithms to make recommendations, but they were programmed by humans – and those humans can instruct those algorithms to reveal exactly what processes they use to arrive at their conclusions.

AI that meets compliance requirements

Transparent AI systems offer a full trail for auditing of investments – the kind of auditing institutional investors are required to supply – with information supplied for each element of an investment portfolio. Investors will thus be able to understand the logic behind each signal, and how they can benefit the institution’s portfolios. Not all predictions will pan out – but at least investors will be able to clearly explain why one investment succeeded, and another didn’t.

Transparent and understandable AI is something that investment firms should consider also in light of possible regulatory requirements. Government regulations on issues like money laundering and insider trading have become significantly more stringent in recent years, and investment managers, especially at bigger institutions, are more likely to be asked by regulators to explain their investment strategies – and the likelihood of that happening may be even greater for managers who use advanced AI. With transparent AI, managers will be able to quickly and efficiently document their investment strategies, providing assurance that, despite the fact that they made significant profits, those profits were obtained without violating any regulations.

With that kind of system, investors can take full advantage of what AI has to offer – and they can be sure that they will be able to explain to those to whom they are responsible exactly why they did what they did. Investment managers will be able to leverage the power of AI to prove and capture the alpha in their investment theses – leading to a new paradigm for investing, where managers are able to make more intelligent and safe choices – backed by powerful algorithms that help them succeed. Such an approach will make AI into a truly transformative technology for institutional investing.