Why AI Won't Replace Stock Pickers - Jack Woodruff, Candlestick & Vinoo Ganesh, Kepler
https://www.youtube.com/watch?v=y2dyD6slKxE
AI Engineer · 18 min · 236 sentences · sentence-level transcript from YouTube's caption track; each line begins at the time shown

[00:01] [music]
[00:12] I am the founder and former CIO of Candlestick Capital Management, a long-short equity hedge fund based in Greenwich, Connecticut.
[00:20] Before founding Candlestick, I was a partner and portfolio manager at Surveyor Capital, which is a division of Citadel.
[00:31] Very cool.
[00:32] Thanks, Jack.
[00:33] So, the reason we did this this way is we've had this conversation before, and I think this is one of the most enlightening folks to chat with, given that Jack has spent his career
[00:43] picking stocks in an environment that's changing more than it ever was before.
[00:47] So, we'll treat this really as a as a fireside chat.
[00:50] So, Jack, paint the picture for us.
[00:53] What does fundamental investing look like in the age of AI, say 5 years from now?
[00:57] Well, it's it's hard to say.
[00:59] Like, long-term stock picking, I don't believe it's really going to change much at all.
[01:06] There's always been a mosaic, and there's always been alpha creation by finding a divergent view.
[01:13] So, that isn't going to change at all.
[01:16] I think that that still will stay the same.
[01:18] Now, how people synthesize all of the information and where table stakes go to and to learn in terms of like what everybody has, and then how you synthesize that data
[01:30] is going to change completely.
[01:33] The roles the traditional roles of say a portfolio manager or an analyst are going to be completely thrown upon their heads, and I still think it's like too early to say what that's all going to look like
[01:46] as these tools are just so new, especially to the users.
[01:51] You got a lot of people doing a a few things, but I it's just not clear to me that it's like we're there and oh my gosh, it's going to be these new things.
[02:01] Like we have we have a bunch of other things to kind of talk about, but um I think it's going to be super fascinating to to to watch how it all plays out.
[02:10] Very cool.
[02:11] So, at Citadel, you had 18 people covering 400 stocks, and at Candlestick, you had 30 stock pickers.
[02:17] Could one great analyst cover all 400 by 2030?
[02:21] So, the biggest issue is is is humans are still the bottleneck.
[02:26] I mean, like at the end of the day, uh you have to get the information into your head.
[02:33] So, it's really all about sort of how man and machine can work together.
[02:38] Um investment management is a time management exercise.
[02:45] What I mean by that, if I spent all of my time working on one stock, that'd be great.
[02:51] I'd probably know better than anyone else in the world, but like if there wasn't anything to do in that stock, I'd be kind of there there isn't much to do.
[02:59] Well, if I do two, well, then I have half as much time, and and so forth until I go.
[03:04] So, you start getting diminishing returns on how much time you could spend on any particular stock.
[03:12] So, you know, what can AI do?
[03:15] Can AI help me cover more names?
[03:18] Perhaps.
[03:19] Um maybe it allows me to raise the table stakes on what I just kind of like have as a general knowledge, um because that work used to take me hours to prepare to get me ready for what it was, and now maybe the, you know, an AI program can, you know, help me prepare a lot of that
[03:38] work in 5 minutes, where it would used to take me 6 to 8 hours.
[03:42] Um So, that allows me to do other things.
[03:45] Again, like it's a very competitive industry, so it's not like if I get rid of all these people, or if I have like five analysts, whereas before I had 30,
[03:53] um yeah, maybe those people can do more, but now all of a sudden this opens up other time this opens up time that I could spend doing either more proprietary
[04:04] research, channel checks, more sort of like industry backdrop work because again, like you're going to need to find these divergent views.
[04:13] So, it's like if I do the exact same thing I was doing before and it's like, "Hey, wow, I can get my work done in 2 hours."
[04:20] Well, someone else is going to be still working 16 hours a day trying to find like that extra thing that's going to that that that's going to make you money.
[04:30] Because, you know, just doing this stuff is is obviously like what we used to pay an analyst to do, you know, a year or two ago is not what we're going to be paying an analyst to be doing, you know, in a year.
[04:45] Very cool.
[04:45] And on that, what's What is something today that a junior analyst does that you think will be absurd if they're still doing in 5 years?
[04:53] Well, it's funny.
[04:54] Like it's like, "Oh, you're you're not going to be reading 10-Ks.
[04:57] You're not going to be reading, you know, transcripts."
[05:00] I actually don't believe that to be true.
[05:02] Like if you're truly close
[05:04] [snorts]
[05:04] to a name, you're going to still need to like process a lot of that information.
[05:09] Now, you might be able to get like summarized versions that can cut down the time it takes you to do those things, but you still need to get that information sort of into your head.
[05:20] Like we used to have a lot of analysts that would be going through a lot of like sort of like big data and things like that.
[05:26] I definitely think that that can sort of be outsourced and that'll be super easy to you know, I would be surprised if if a lot of these guys aren't necessarily doing that stuff.
[05:35] But what I believe the AI is going to do or the agent or whoever it is is going to be pushing you in the right direction to work on the right names.
[05:49] Like I said, this is time management.
[05:51] It's all about how do I know to work on the right things?
[05:54] And that's where the winners are going to is going to be is is the people who can figure out how to use these tools to get you to spend
[06:01] the right amount
[06:02] [snorts]
[06:02] of time working on potentially finding like a divergent view.
[06:08] Very cool.
[06:08] And on that big data side, so analysts spend a huge portion of the week finding, clearing, and checking data.
[06:15] Cleaning and checking data.
[06:16] When that goes away, do they get smarter or just busier?
[06:20] I I like
[06:22] [laughter]
[06:23] like anything.
[06:24] Well, now that we have extra time, we'll be looking for more data.
[06:28] We want more information.
[06:30] Uh I need to be doing more proprietary research calls.
[06:33] I need to be doing more industry builds.
[06:34] I need to be doing more and more things.
[06:37] And it's just going to raise the bar on the amount of work that everybody's doing.
[06:41] And so, when everybody's doing all the exact same things, that just becomes the table stakes.
[06:47] It's like it's just automatic.
[06:48] I have to have it.
[06:49] I can't I you know, I I I'm not a I'm not competing.
[06:54] And then it's all about finding that next layer of information um about how there's going to be a divergent view.
[07:00] It's like that's that's it.
[07:02] That's part of the mosaic.
[07:03] And it's not like you're just finding this one data point, and that data point's going to be the thing.
[07:07] The investment mosaic is all about finding three or four sort of data points that all sort of are lining up that aren't necessarily coming from the same place.
[07:15] And that's generally, I think the widely accepted view of how alpha is is is created.
[07:21] Very cool.
[07:22] You have a pretty interesting background.
[07:24] You started off as a CPA.
[07:26] Would you ever sign off on a number a model produced without seeing where it came from?
[07:31] No, no, no. Absolutely not.
[07:32] I mean, that's like actually the most exciting thing about what's going on at Kepler is just our ability to, you know, audit every single number.
[07:40] And, you know, when these tools start running amok and you you Whenever we made mistakes, cuz we had humans entering these things, it was very rare, but like the most important thing for us is that we got the numbers right.
[07:54] But still, mistakes were made from time to time.
[07:56] And like, you know, millions of dollars could be lost off of a single sort of like data entry mistake.
[08:02] Um It's funny like I there's no way we're going to let AI, you know, tell us exactly what to do, run trading algorithms, all these kinds of things, if you can't trust
[08:12] the information cuz then all of a sudden it could start it could run a muck it could be anything.
[08:16] Like the the the accuracy of the numbers and your ability to audit it um is is absolutely the most important part of this entire process of being able to sort of like trust some of these things and
[08:31] you know, I know a lot of people that are using these tools today and so they'll ask, you know, Claude or somebody to to build help them build a model and stuff and then they still have to go through and audit the numbers to make sure that that they're all there.
[08:44] So it's like are we really
[08:45] [snorts]
[08:46] saving that much time by automating this?
[08:48] It's it's I think everybody assumes oh it'll get better it'll get better but again like when you know you can actually audit everything it's I think that's going to be just a a huge
[08:57] important factor.
[08:58] It's it seems obvious
[09:00] Yeah.
[09:01] and easy but uh it's apparently according according to the
[09:05] No, it's not so.
[09:08] Awesome.
[09:08] So for a lot of the AI companies here, Jack is the buyer.
[09:12] So like this is the person you convince at hedge funds to buy software.
[09:15] And so I'm curious, what part of your investment process would you refuse to let an AI vendor see?
[09:21] So I'm the wrong guy to ask this question because like this is not my like security of of data and you know, like this is a total CTO question and so I have
[09:35] a great CTO and like he's the guy who would let handle it and and he would sort of like be telling us.
[09:40] I would be very nervous about any sort of internal um data.
[09:45] Like listen, you you don't want anything to get out as long as you can kind of I don't even know what the terminology is I'll sound like an idiot in front of all you engineers but like,
[09:54] you know, I don't want my stuff to go there.
[09:57] Like
[09:57] [laughter]
[09:57] you tell me whatever that's called.
[09:59] Data security isolation.
[10:01] Yeah, exactly.
[10:02] So I would want to make sure all of that um you know, is is is very critical.
[10:08] And I mean I think to like where we're going is like so much of like what needs to happen first is I I think you know, Venu and I have been calling it like almost like this is like 1.0.
[10:19] Is it's like we're working on the automation part.
[10:22] We're going to get all these tools in there.
[10:23] We're going to have all the data sort of built up.
[10:25] But then we can actually like really start doing interesting things and we can use the tools to help the analyst to say, "Here's where you should be spending your time.
[10:35] This is where it gets more interesting."
[10:37] Um because again, it's all about It's Look, investment management isn't about doing earnings previews.
[10:43] It's not about building DCFs.
[10:45] It's not about that.
[10:46] It's making money.
[10:47] And so, you to make money, you need divergent views.
[10:51] And to find divergent views, these are tools that help you identify those types of things, but they're not the output.
[10:56] Um so, that's going to be like the real critical thing.
[11:00] So, anything, you know, that you you feel like becomes proprietary in in your way of making money.
[11:05] Now, one of the hardest things in investment management is you know, competitive moats.
[11:11] Um like finding some sort of arbitrage.
[11:15] They generally last pretty short.
[11:17] You know, it's all very smart people, motivated, highly compensated.
[11:21] And when someone else is making a ton of money, other people sort of figure it out and arbitrage gone.
[11:29] Very
[11:29] [snorts]
[11:29] cool.
[11:30] Once every fund has the same models, does edge come from better data, better questions, or better judgment?
[11:36] I think all of it.
[11:37] I mean, like it it's everyone is ultimately going to have all of the same information.
[11:43] So, it's all going to be about how you prompt it, how fast you're able to sort of like, you know, combine man versus machine.
[11:51] Like the way that we used to always think about is like we did this enormous amount of work that I think can actually be automated now.
[12:00] Um that it didn't tell us, "Hey, go buy this stock."
[12:04] It said, "Hey, this is something you should go spend a lot of time on.
[12:09] This is a project where you need to go and like really educate yourself.
[12:13] And that's when you go and you find something because not everybody is looking at that.
[12:18] Everybody else is looking at all this other stuff and everybody has that sort of thing, but you're looking for that sort of needle.
[12:24] You're saying like this might be interesting and if I can get myself if I can get more conviction around this and if this happens, we're going to make a lot of money.
[12:34] And so that was it.
[12:35] That was the purpose of all of the field work.
[12:37] And so like what's going to happen is all those types of things.
[12:39] And so it's going to be about how people utilize the tools to really identify these types of things that we can become special projects.
[12:49] If AI can just do it, then I don't know.
[12:51] I mean it's you know what I mean?
[12:53] What do we need us for?
[12:54] So um but I think it's going to be you know a way to use the tools to um to to to figure out where you need to do more research or find something sort of like different to do because there's an unlimited amount of things one can do.
[13:12] And so it's like you say, "Oh, we'll just get great.
[13:14] AI can do it all."
[13:15] And it's like that's not necessarily true cuz it's like we got to find the right things and it's all about identifying those things.
[13:21] It's about identifying them fast before everybody else.
[13:25] Very cool.
[13:26] So you launched a very successful hedge fund.
[13:28] If you launch a fund in 2027, what's on your build list versus your buy list?
[13:33] Well, I'd obviously buy Kepler.
[13:35] I mean that's
[13:35] I agree.
[13:36] I agree.
[13:36] [laughter]
[13:39] Um
[13:40] [sighs]
[13:41] [gasps]
[13:42] It So this one is like I'm not sure yet.
[13:47] Um I Obviously, I know there's going to be a lot of changes in people's roles.
[13:57] I still need senior stock pickers.
[13:59] I may not need as many analysts, but I might need other people with different skill sets um that might replace like an analyst type of role.
[14:10] Investigative journalists, people who can do like really deep project work.
[14:15] Um, you know, maybe different sort of data engineers.
[14:19] Like it's I'm not 100% sure what it's going to be.
[14:22] It's going to be very different.
[14:23] The way that our uh you know, kind of structure was it was a pyramid and it was like we had a ton of analysts presenting a ton of information
[14:32] to senior stock pickers.
[14:33] Senior stock pickers presented information to a couple of PMs and and the CIO on top.
[14:38] And so that was it.
[14:39] Like I kind of see it being a little bit more vertical.
[14:42] Like I still need all of the I want the the producers.
[14:46] I I want the the makers, you know, the people who are really there to make the money.
[14:50] It's I don't know if I'm going to need as many of the preparers.
[14:54] Um but I also like might want them because I might need more I might be looking for more proprietary information to help me fill you know, feed the beast or feed the machine
[15:06] because that's those are going to be things that hopefully allow me to you know, find you know, like these divergent views that I'm looking for.
[15:14] Very cool.
[15:14] So who in your opinion does AI help more?
[15:17] A five-person fund that can act like 50 or a 5,000-person fund that can act like five?
[15:22] So it's funny.
[15:25] The one thing that everyone needs to like get their heads around is every hedge fund has a different strategy uh whether hedge fund, investment fund, like any anybody who's investing
[15:37] [snorts]
[15:38] and they have a different mandate.
[15:39] Some people are I am going to turn over my book 9,000 times a year and it's just like day trading and constantly high frequency and some people are like I buy and hold for 5 years.
[15:50] And so it's all sort of about like what is your strategy?
[15:57] Everyone's going to be able to do more with what they have.
[16:01] And so but what's that going to do?
[16:03] So if I take a five-person per uh team I would imagine because they only have five people there's slower idea velocity so it's probably going to be a more concentrated
[16:11] type of strategy.
[16:13] Um if I have more ideas to choose from, then in theory my hit rate should be better.
[16:23] So, if I have like, all right, I cover 100 stocks, uh I got to have 20 names in my portfolio, so I'm picking 20%.
[16:30] Well, if all of a sudden I'm covering 200 names, then I've got 10% of them needs to be in the 20.
[16:36] If I'm covering 400 names, I only 5% are going into that book.
[16:41] And so, the chances of me being right, but the research for those 400 has to be just as solid as the research on the 100.
[16:48] So, the those are the types of things.
[16:50] Now, I'm a 5,000-person firm or whatever, and uh that's a lot of people.
[16:54] Um there's obviously there's probably you're a multi-strat, you're you're doing every asset class under the sun, so it's it's very different.
[17:02] But again, idea velocity and the number of good ideas you have in your pipeline is paramount for an investment firm.
[17:12] Like, when something works, the first thing you want to do is you want to get it out, so you can put something new in that works.
[17:18] And so, like when when everything's going perfect, every time something works, I replace it with a new thing, and then I get rid of it, and then I replace that with a new thing, and then we and that's how we, you know, build alpha and build build returns over time.
[17:30] Um and that's, you know, what we were trying to do.
[17:32] So,
[17:35] [sighs]
[17:35] Very cool.
[17:36] What's an AI idea that everyone in finance is excited about that you think won't work, and the one no one's talking about that you think will work?
[17:44] Well, I wish I knew the the thing that nobody's talking about, cuz then I can make some money.
[17:48] But um the the thing that everybody's sort of talking about is this no-analyst hedge fund, you know, oh, I'm just going to have all these AI agents.
[17:57] What I do believe is that they will be able to replicate a fund that existed a year ago.
[18:05] It's just not going to be as good as what today's hedge funds are going to be next year.
[18:11] And so, I just don't think it's ever really I don't think you're ever going to really be able to take humans out of the loop.
[18:17] Um and when you can take humans out of the loop completely uh on a fundamental and a you know, quantitative funds have taken human out of the loop for for decades.
[18:26] So, that's that's nothing new.
[18:27] But, from a fundamentals perspective, when you can take the humans out of the loop, then well, then it's just going to be whoever's got the most compute, the most
[18:35] the most information.
[18:36] So, I just don't think that that's how it's going to go.