"The Cognitive Revolution"
01:01:45 · Marc Bhargava, Nathan Labenz
Roll over a significant amount and be part of an alternative to that sort of PE consolidation. And we think that really resonates. Honestly, I think it can resonate for not just 10 or 20%, but for a majority firms. And certainly, every year, we're seeing it resonate for more and more. [01:02:00] S0: Okay. So here, I'm gonna push you on the... Just how crazy does this all get. But maybe we can start with... You went through 70 industries, you know, winnered it to 10 based on where you could see a path to 30% automation. What are the things that can't be automated and, like, what are the barriers? I mean, to some degree, obviously, like, operating in the physical world, like, we don't have robots yet. We do have, you know, quite a few humanoid companies that are, I think, on track to maybe change that in the not too distant future, but clearly, it's not here yet. Then we have this notion just from the AI companies that there is gonna be the drop in knowledge worker, and it seems there that one of the big barriers is context, like lack of context or lack of ability to handle enough context, etcetera. How do you taxonomize what can't be done? And and do you have a map for, like, what the AIs specifically can't do that makes those things not yet accessible?
01:03:02 · Marc Bhargava
[01:03:02] S1: There are kind of two bigger buckets here. So one is what you already alluded to where, you know, a lot of the automation today is more in these continuous tasks that I spoke about, like data entry, analyzation, creation of presentations, those sort of things. It's not in, for example, you know, maybe plumbing and roofing and HVAC and a lot of like in person, work, firefighting, etcetera. And so there is a lot that still happens in the physical world where there might be some automation on maybe doing routing, scheduling, payments, all of that, but it won't meet our 30% criteria of like, look, this automation will really transform how you are a firefighter, for example, at least not right now. So one bucket is that. A second bucket is we don't want to go and buy up firms that are have high churn. So you know, when we come in and we do AI automation, if you think of something like the digital ad agency or PR agencies, lots of folks use multiple ad agencies and multiple PR agencies. So if we were to go and buy one and start transforming it with AI, if there are any hiccups along the way, they might just ditch us because it's a very kind of using multiple providers and it's very high churn. And so we also... Another big criteria is we're only doing this in industries where there are really good business dynamics, where the customers have sticky contracts, maybe they're multi year, there's very low churn. So when we come in and we start doing AI automation, it's not always a 100% smooth. And so we have to have the benefit of at least six months or a year to kind of implement these things for our base workers and for the end clients as well. So those are two things that really have struck out a lot of different industries where either there's a big real world component or business economics around churn are are kind of dangerous. And so those have been two areas at least recently that have sort of flagged it so that it's difficult to go in that industry. And then a third of course is looking at how good is the AI as well. And so, you know, two years ago you wouldn't be able to use AI to automate a lot of what we're looking at today like in insurance and even looking at insurance underwriting like that was just AI totally wasn't there.
01:05:07 · Marc Bhargava
So there are also other industries where, oh, maybe one day AI could do those tasks, but the current models are worse than humans doing it. So you're not adding any real value there.
01:05:17 · Nathan Labenz
[01:05:17] S0: So do you not expect to see this drop in worker? Because, I mean, it does strike me that there's, like, several data points that I see pointing to big tech takes all. A friend of mine coined the term the big tech singularity. And I think that is basically my mental model of what's gonna happen. Not necessarily, of course, like, that they'll own every customer relationship or whatever. You know, when you... With Google and Facebook just on the ad platform, of course, they had, like, all these, you know, sort of local agencies and sellers and people, you know, kind of translating the platform value prop to the local business. But talk about a business that's had a lot of margin compression, you know, those facilitators of, oh, I can do Facebook ads for you, mister local business owner, that's been a really tough spot to be in. And, you know, in medical, for example, we've got, like, multiple studies that have shown that the AIs can diagnose better than the doctor and can even like prescribe better than general practitioners at least these days. There's been some that have shown that like the hybrid, you know, there's always this like refuge that people want to take and well, it'll be the doctor and the AI that'll work together, but there's been some suggestive evidence at least that like the AI is actually performing the best, then comes the hybrid, and then the doctor alone is worse. I just also saw a data point from Harvey the other day where they put out a a stack ranked leaderboard of... On their own, like, big law bench... Benchmark that their own fine tuned model has kind of fallen behind, like, eight other foundation models that just seem to be winning on the bitter lesson and generally, like, scale continuing to pay off. There is still this issue of like the drop in knowledge worker doesn't quite exist because it's hard for them to absorb all the context, but that to me seems like something that probably gets solved. I guess I wonder like, do you think that just never gets solved? Or if it if it does get solved, how do we not end up in a situation where it's like, okay. Now we're just dropping in these AI knowledge workers into this environment. And, you know, again, you sort of have a... Now you have a very literal complement versus substitute dynamic going on.
01:07:28 · Marc Bhargava
[01:07:28] S1: Yeah. I think for us, you know, it'll be a lot of individuals will manage teams of agents. And so then you can say, is that replacing, you know, the the folks they used to manage? Not necessarily. Like, I do think that the overall economic growth and GDP output can grow pretty significantly. And so we'll kind of have this land of abundance where people are still doing jobs but with teams of agents below them, but there are way more jobs to be do... Done because it's such a larger global economy. And I do think that pretty much every technology shift we've seen has begged the question of isn't this going to mean less jobs? And it has meant less certain jobs, but the overall economic output has grown. And with that, there have been kind of newer jobs and different opportunities as well. So it is very hard to predict and lay out exactly, like, what's the overall economic growth? You know, what does that mean in terms of new jobs? Which jobs are lost? Which jobs are gained? I think we absolutely have to be mindful of that. But at least as far as we're seeing in the cases of hybrid services, software, AI enabled, you know, the strategy we're doing is not around really replacing jobs. It's freeing people up to do other tasks. The number of tasks that can be automated certainly are growing more and more, but then there's other opportunities, especially like client to client services and lots of businesses. Need to have a relationship. Even the doctor example or the nursing example, you know, in nursing, we incubated a company called Hippocratic, which is an AI native nurse that we work with a lot of healthcare systems to get proprietary data to be able to help build this company with the founder and it's part of our creation strategy. There could be 10 times more nurses in this country than there are. Right now we don't do preventative medicine calls. We don't follow-up after surgery post six months or nine months. There just could be a lot more nurses and there's an absolute shortage. So many of these industries that we're going after, there's a shortage of people. We need to give them the tools to match that shortage, and I think a lot of that will be AI enabled agents, for example, working under them. And so that's the reality of the view we see on the ground today.
01:09:29 · Marc Bhargava
Can we debate if, you know, robotics is certainly getting better and enable robotics is getting better? Are we gonna eventually go to this other world, etcetera, etcetera? I mean, that's a very hard question to ask, and there's all kinds of technology and automation improvements. I would just say, me personally and General Catalyst, we are very focused more on what are the opportunities in front of us today. And we have been really mindful of the companies we back. We're a large investor in Anthropic. We think they do an excellent job of thinking about the risks around AI and AI safety. And so, you know, responsible innovation, AI safety, those are really key to our mantras and the companies we're investing in. And we think like every other technology, as we adopt this, there will be challenges. It's the mindset you bring to those challenges.