Written with AI: Why AI Won't Replace Fund Managers & Equity Analysts Anytime Soon
My exciting experiences with incorporating AI into equity research thus far.
The first half of this article was handwritten by me. Scroll down to the second half of this article to read the AI-written part in the following section.
I’ve been experimenting with incorporating AI into my equity research workflow, and have been very satisfied with my progress thus far. As you can see in the picture above, this was me attempting to use AI to extract the EBITDA per sqft performance from a particular company’s annual report — without first requiring hours of data entry into Excel (as it usually does). While the initial question didn’t yield useful results as can be seen above, anyone with buyside experience will know how to use the correct prompts to prod out the desired information from the documents, as long as such information exists inside.
Honestly, I see this as being a game changer for equity research, and am excited to continue learning how to further incorporate AI into my equity research workflow. I’m keenly aware that I have barely begun to scratch the surface of AI’s potential to augment equity research, and am looking forward to mastering it over the coming year(s). This should initially manifest as greater research output — and hopefully even more creative and higher quality research findings and presentations later on.
Perhaps an illustration is in order to demonstrate how useful AI can be to equity analysts and fund manager. As someone truly passionate about markets, one thing that has always pained me is that I regularly do much more research than I’m capable of communicating on this newsletter. This is primarily because typing and editing articles is a chore which does not add much value relative to the time it takes — which could instead be spent on doing more research. As a result, I have had to be very selective about what kinds of articles I published in the past, despite easily reading 10x more material and synthesizing them into actionable market insights in my head.
However, I’ve recently discovered a new AI workflow which should allow me to pump out more such articles — perhaps even on a daily basis, time permitting. This rudimentary hack involves: 1) recording myself opining on future market trajectories in a disorganized, stream-of-consciousness manner, 2) using Whisper AI to turn the recording into a transcript, and 3) asking AI to turn my disorganized thoughts into an organized article.
There are some incredible time-saving benefits to doing this, as opposed to typing and editing an article as usual. To illustrate, I shall use an example of writing an article about macro research. For context, macro research usually involves a sprawling number of data points across multiple different fields and branches (e.g. making interest rate forecasts requires insight into monetary policy, fiscal policy, liquidity flows, forex and geopolitics). As the conscious mind typically only has enough mental bandwidth to describe one chunky macro topic at a time, what usually ends up happening when I want to write a macro research article is this: First, I have to pour out all my thoughts in a disorganized manner to get everything out on paper; then I have to rewrite the entire article to organize it into a coherent format with logical flow that readers can readily follow. As you can imagine, this involves a ton of editing time.
Using the AI-assisted approach as described above, what I’ve been able to do instead is to let AI handle the second step of reorganizing all my scattered thoughts into an article with logical flow. I still do the regular first step of spilling my brain out into a recording; which is then transcribed into something legible by Whisper AI. Subsequently, I can feed the entire disorganized transcript into ChatGPT or another similar AI tool, and then simply ask it to rewrite it as an article. While the article still possesses some of the tells of AI, it actually helps significantly to reorganize the logical flow of my disorganized thoughts into a coherent article that readers can readily follow. And when discussing macro takes you down multiple rabbit holes at a time, it really helps to have AI build an underground highway through all of them at once for your readers — all in one click.
Obviously, this does not take away the need to possess all that macro insight to begin with in order to produce that initial recording with key information. However, it does help tremendously in making it extremely easy to edit that information into a consumable format. Going forward, I will be experimenting with several different AI-assisted approaches to get my thoughts on markets out as speedily as possible, which should substantially increase the value of this newsletter. As a former fund manager, I regularly discuss my cursory market insights with friends in a private capacity — hopefully AI will make all these years worth of accumulated market insights accessible to thousands of others in a more public capacity as well. And fret not if you’re not interested in macro — I’ll be using this method to convey more lessons from Buffett’s career and Value Investing to you as well!
As an example, the article below was written by AI using this method (with some editing). Don’t forget to check out the separate macro article linked at the end too!
Written with AI: Why AI Won't Replace Fund Managers and Equity Analysts Anytime Soon
The article titled "Why AI Won't Replace Fund Managers and Equity Analysts" presents a compelling argument about the limitations of artificial intelligence in the context of financial analysis and fund management. While the potential of AI as a valuable tool for professionals in the industry should be acknowledged, it is not likely to replace human experts anytime soon.
One of the main reasons for AI's inability to supplant human fund managers and equity analysts is its reliance on the quality of input it receives. AI can perform tasks such as calculating operating margins for a specific fiscal year, but it requires precise prompting to do so, which in turn requires deep industry knowledge to perform. For instance, when tasked with finding the operating margin of the company for FY23, the AI initially failed until it was directed to the correct page containing the necessary information. This possibility may not have occurred to users with more limited experience in financial statement analysis — and naturally, more advanced requests will benefit significantly from deeper prompting, which may require deeper industry experience. This example illustrates that without specific industry knowledge, a layperson cannot effectively prompt AI to perform complex financial analyses — i.e. the AI is only as good as the person using it.
The following analogy further illustrates the difference between AI and human expertise. AI without specialized prompting can be likened to a stone, which can be used to break something but lacks cutting precision. In contrast, someone with advanced industry knowledge can prompt AI in such a way that it acts more like a knife with a very sharp edge, capable of making precise cuts. This analogy underscores the idea that while AI can augment base capabilities, it cannot replace the nuanced understanding and skill set of a professional with advanced financial knowledge, in the context of financial analysis and fund management.
If AI were ever to reach a level of artificial general intelligence (AGI) where it could brute force solutions to any analytical problem, it theoretically might be able to replace human intelligence one day so as to render equity analysts and fund managers ancillary. However, this level of AI sophistication is perhaps decades away — it would basically be the same as saying that AGI could theoretically figure out how to make cold fusion practical, and thus render all traditional energy sources worthless overnight. The current iteration of AI, while useful, still requires significant guidance to produce the desired results.
In conclusion, AI's current role is to assist rather than replace human fund managers and equity analysts. The quality of AI's output is directly tied to the quality of the input it receives — and without professional industry background and knowledge, it is challenging to ask the right questions of AI. Therefore, while AI can provide advantages to laypeople by enabling them to search for information which previously might not have occurred to them, it cannot match the level of inquiry and analysis that a professional with advanced financial knowledge can achieve. At the end of the day, AI merely augments the base research capabilities of the equity analyst (i.e. helps everyone stand on tiptoes) — which quite counterintuitively, might actually end up widening the gap between laypeople and industry professionals.
AI, as it stands, is a tool that enhances capabilities but does not possess the depth of understanding necessary to take over the roles of fund managers and equity analysts. AI's utility is undeniable, but its ability to replace human expertise in financial analysis and fund management remains limited.
Very interesting what AI can do. Great work