Beyond the hype of artificial intelligence: What every investment researcher needs to know

Artificial intelligence: the new technology wave with no end date in sight. Or so marketers would love for you to believe.  In actuality, the hype surrounding AI has conflated reality with science fiction, making it difficult to determine real business value even when it is plentiful.

Nearly all industries are analyzing the potential of artificial intelligence and machine learning, with many seeking to identify AI experts to help lead the way. Amidst all the hype, many can’t help but wonder: Where can artificial intelligence be most useful today?

To answer that question, it’s important to note that AI will always be most effective at performing narrow, specific tasks, at least for the foreseeable future. AI can go much deeper on a specific subject, analyzing more data than any human analog to observe a specific outcome, but it can’t always execute on the big picture. If you are aware of the strengths of AI, but also know its limitations, you can make the most appropriate use of what AI and machine learning have to offer.

Consider investment research — a field that requires both expert comprehension on a variety of data sources, as well as quantitative analysis and risk management techniques to execute comprehensive fund strategies. This is a field primed to capitalize on AI technology, but not when implemented naively.

While there’s much value to be gained, it won’t happen just by shoving all the data into a table and pointing a neural network at security price. AI can easily feel overwhelming when trying to determine real, valuable use-cases. Here’s a look at the three best integrations for AI in investment research:

#1: Processing large volumes of qualitative data and language sources

Traditionally, investment research requires a great deal of qualitative assessment on various sources like earnings, press releases, market sentiment and government filings. And now, much of this data is easily attainable, and thus accessible to machines. While the volume of information is significant for human teams, there are ways to augment and incorporate machine learning to expedite grueling analysis.

Natural language processing as an AI technique has developed so much so that extracting text from documents and transcripts, and searching and tagging that data has become commonplace. And now, even the reading process can be expedited by creating intelligent summaries of documents and then noting the most pertinent and relevant information.  

At Trill, we manage to combine all of our researchers’ qualitative sources into one place where they can manage updates, see quick summaries and high level overviews of their data, and ascertain the sentiment of those documents at a glance. This allows researchers to work twice as fast as before. The use of AI technology not only makes everything readily available, but speeds up the information and decision-making process.

#2: Projecting forward-looking financial line items and earnings estimates

This process is already quite quantitative and often includes many complex Excel models and statistical analyses, which are used to identify undervalued companies and project forward cash-flow estimates. But, with machine learning techniques, this task can be expanded a hundred-fold, allowing for broader data inclusion, and multi-company coverage to identify subtle but real underlying relationships.

At Trill, we’ve seen investment researchers put this into action, using deep neural networks to project future company earnings one year into the future with a mean squared error of only 0.33.

Though the quantitative data is often accessible, the relationship to financial projections is not always obvious—which is where AI technology steps in to fill that gap.

#3: Projecting forward-looking risk exposure at the company level

Another highly quantitative field, risk management also includes many complex statistical analyses including covariance analysis and risk modeling/attribution. However, with machine learning techniques, we can more subtly project the future importance of underlying risk variables. While most risk attribution is backwards looking in nature and projected forward polynomially or linearly, leveraging an appropriate neural network architecture allows us to determine the non-polynomial, forward-looking exposure of financial statement and macroeconomic data towards stock performance.

At Trill A.I. we are projecting the forward-looking risk exposure at the company specific level for all US and international equites. Top quantitative hedge funds are leveraging our data today to more accurately model performance risks and build more comprehensive trading strategies.

How do you evaluate which factors your holdings and trades will most exposed to going forward?

The value of AI simplified

Investment researchers who choose to harness the power of AI to analyze give themselves a competitive edge. And, as AI begins to be utilized more and more, they allow themselves to be ahead of any imminent learning curve.

So, ask yourself:

  • How much more could you do if you offloaded elements of your investment research to AI-powered technology?
  • How might the research and recommendations you provide improve with additional data analysis?

We can’t create more time in the day for ourselves, but we can scale our capabilities by leaning on AI and putting it to work under narrow, specific conditions that result in big wins.

Image: Pexels