There are days when you’re staring at your Bloomberg Terminal’s screen, refreshing your inbox, or constantly pulling down on your screen to refresh the newsfeed on your phone. That’s earnings season for you: companies you cover reporting on their quarterly performance, their competitors, and their mothers. Everyone’s reporting and spewing out an unfathomable amount of information. Meanwhile you need to scan the releases, spread the numbers on your Excel tracker and synthesize all that information and pray you can answer the one question that matters: are these results good or bad? And when you’ve settled on an answer, you have to conjure a coherent narrative in a written note and send it out to your email list; all within the hour.
The job gets easier the longer you do it, you get better at weeding out the noise and picking out the material commentary in all these releases. The only problem is every other analyst is also exceedingly good at this very task. Research, synthesis and writing become table stakes in an industry where you’re continuously trying to differentiate.
This is where more context can make a difference. When the current conversation is around big tech’s elevated capex spending, it is valuable to revisit capital spending cycles in other industries, understand their drivers, and their constraints for a clearer picture.
Filings as fingerprints
I previously discussed the role of AI in structuring company disclosures and clustering them around themes:
I want LLMs to read hundreds of filings for all the listed tech stocks over the past four quarters and produce a structured summary that highlights what’s discussed, the sentiment, and why it’s being raised. That suits my needs — finding themes in filings — though the structure could be adapted to whatever an analyst cares about.
NanoSlopt just reported its earnings and they commented on their increased data center construction activity. This is where themes come to the rescue; you can traverse your filings index and find a related theme: “infrastructure investment”. That theme encompasses filings from companies across a wide breadth of industries; a telecom company’s investment in 5G infrastructure, capacity additions at an automotive company’s factories, or a retailer’s investment in warehousing.
You have essentially shifted your analysis from a vertical look at the company’s past performance and direct competitors in its domain, to a horizontal search grounded in a common theme: expansionary capex
A system for finding analogues
You now have tens of filings from companies in industries you do not cover. You can stop there and just Ctrl-F your way through deciphering the investment patterns of these companies and pattern match with NanoSlopt. But you have one hour to wrap your earnings take. The good news is AI is good at acting autonomously and following instructions; reading text and taking actions. So instead of the manual readthroughs, you can feed all those filings to an AI-powered system that gathers context and finds cross-industry patterns. You can instruct it to research like you would; typically asking it to look up associated risks and expected outcomes of such an investment.
A semi-autonomous system following instructions in natural language can fully replace research workflows designed to fast track you to an early retirement.
I created a miniaturised version of that system, but the real bottleneck is the data. For such a system to operate at a competent level, it needs access to thousands of structured filings that it can refer to. I’m working on it.