The new age of Research Discovery is not just about GenAI

Everyone in the Investment world has been excited about the potential of Large Language Models since ChatGPT rocketed them into public consciousness in late 2022.

Why?

Like any industry, LLMs potentially provide significant productivity gains for the investment research business. Some might even go so far as to say that they could replace investment analysts, or at least, some sections of their process. Wild conjecture at this point. However, from a more practical perspective, there is a great deal of optimism that LLMs might solve one of the enduring problems of the investment research business; Research Discovery. Solve this, and major productivity gains will ensue. Who knows, it could also be a return enhancer.

And what is the problem exactly? 

Simply put, the sell side (and wider research provider sector) produces too much content for the buy side to process, synthesize, and consume, meaning they only gain fractional utility from it for the money spent. Just consider the long running trend of exceedingly low document open rates as the obvious empirical evidence of that problem.

Over the years, portfolio managers (the main category of research consumers) have devised their own methods to discover written research. These are often imperfect, relatively manual processes that consume valuable time. It’s akin to doing old fashioned gold panning, versus today’s sophisticated, scientific and industrial scale mining techniques.

How can a PM find those two or three genuinely useful nuggets of information or analysis within those thousands of documents? Quite often it may come through sheer luck or determination. It’s no wonder that LLMs have created so much interest and excitement.

Limeglass and Substantive Research are two firms that already provide tools to help clients in the research discovery process. While there’s no denying that Generative AI and LLMs are a significant development for the investment research discovery process, we believe that it is important to realise two things:

1. That most real use cases for Research Discovery can be achieved with other, cheaper, more auditable, less “black-boxy” technologies.
2. That for complex use cases requiring the generative abilities of LLMs, there are some foundational elements that need to be in place before the industry can realise the potential benefits.

These foundations centre on solutions that rely on a ‘’custom mixture of domain-specific tools’’ and human industry expertise. Limeglass and Substantive Research both have compelling offerings for this and have teamed up to share some best practices with you over the coming weeks.

In our next two articles we will ask:

Why is Research Discovery so important? We will delve into the existing inefficiencies in the Research Dissemination value chain and what the future holds for forward-thinking Sell Siders and Buy Siders.

Why is Research Discovery so difficult? We will get all the intricate detail from Simon Gregory (Limeglass co-founder and CTO) on why the product and the existing systems make it so difficult, how people are attempting to solve the problem at the moment, and how it should actually be done.

We will also be hosting a breakfast session in December with demonstrations of Limeglass’ cutting-edge technology, a conversation between Limeglass and Substantive Research, and the opportunity for invited Buy Side and Sell Side guests to ask all the important questions.

Watch this space.