That is true, but it also depends on your work style. For example, I don’t like using these tools for long-form articles because you have to keep prompting them for changes so many times, and then you have to read each finished version completely, that it’s actually faster to write them myself. I write one draft quickly, and then do a single editing pass, and that’s it.

Overall, I’ve found ChatGPT most helpful for 1) Documenting and commenting on code; 2) Writing/assisting with code; and 3) Actual writing, in that order. But I have to admit I never do much “mechanical” writing – each article, case study, etc., usually requires a fair amount of creativity and doesn’t closely resemble previous ones. So I could see these tools being more helpful for other writers.

Thanks for the insight!

Nope. Anyone who’s familiar with the workflow at banks and how they do things internally can tell you why this is the case: 95% of your work consists of “Client X wants us to make Small Changes A, B, and C… or Managing Director Y wants us to change Small Points D, E, and F.” You end up going back and forth on documents dozens of times as people keep requesting small tweaks. All these automation/AI tools are good for rough outlines if you have no starting point, template, or idea of what you want to write. But if you have all those, and your work consists of making tweaks the client wants to see, they’re less impactful. “Prompt engineering” is not an efficient way to make small changes in a 90-page document because you have no idea what changes each new prompt will make, and you don’t want to keep reviewing all 90 pages in each revision.

At most, bankers will say, “Aha, you can draft presentations more quickly now. I’m going to ask you to create or revise twice as many slides!” Oh, and banks are notoriously slow to adopt outside tools. Many banks won’t even let people use simple external macros with completely predictable behavior, let alone platforms where you share confidential client data with OpenAI / Microsoft.

The main risks for IB are still related to macro factors and, specifically for AI, the fact that the cost structures and dynamics of various industries may change quite a bit, reducing visibility in deal volumes/types by sector.

Interesting that makes sense. Thanks for the in-depth answer.

I’m curious if your views have changed a bit in terms of automation affecting the Investment Banking industry after seeing the demo of Microsoft CoPilot. Seems like an industry disrupter.

No, not really, because PE firms also have “clients” (the Limited Partners) and many tasks related to deals and portfolio companies cannot be automated easily. The main use case of ChatGPT in all these fields is to speed up the process of writing/summarizing memos, presentations, etc., which is fine, but it’s only a small part of what you actually do on the job. Another issue is that PE firms are not going to give up all their data to OpenAI or other firms just to “become more efficient” – even services like Capital IQ can’t get the proprietary data with offers to pay a lot for it. So why would they hand it over for free to MSFT, GOOG, or OpenAI?

I think the bigger risk is that AI might disrupt many of their existing investments, make some industries less viable, and completely change cost structures.

I hope to do this as soon as ChatGPT includes a sales & trading module…

(There will be more S&T and HF content on the site this year, but it takes a long time to research and write, so no promises on dates.)

No. It’s still pretty bad for longer/technical content, and it would take me longer to read what it produces and keep asking for rewrites than to write it myself. Also, I expect Google to start analyzing sites for human vs. machine-generated content and penalizing anything that looks even slightly artificial. And there’s already too much low-quality junk out there, so I don’t want to add even more junk to the mix. I think it might be useful for coming up with variations for email subject lines and social media / ad campaigns.