Sequoia's Julien Bek: the next trillion-dollar company will be a software business masquerading as a services firm

ringing the bell at the New York Stock Exchange? I can. I hope it happens. Well, we have our next guest in the Restream. Julian Beck from Sequoia Capital, his partner there. Welcome to the show, Julian. Good to meet you.

Hey guys,

how you doing?

Nice to see you. We also hope we can ring the bell very soon.

Holy holy logo Mog.

Yeah,

holy logo.

Few of those companies.

I do. We're we're partnered with Cisco. Thank you for backing them early, making TVPN possible

through your 30 years ago.

Uh paved the way.

Since this is the first time in the show, why don't you give us just a brief back story on yourself, how you wound up as a Sequoia, what you focus on, what your day-to-day is like.

Sure. Um so, first, thank you both for having me. Uh it's it's an honor to to be there. I know you had a lot of my partners as well. So we're we're very excited

to to be on the show.

Um my name is Julian. I'm a partner on the team. I'm based in London, but it's been a lot of time going back and forth

uh to spend time with the rest of the team in California.

And I started as a founder very very long time ago in high school.

Uh built companies in hardware before uh and so that's why I've been investing mostly in software um for most of my career. Yeah.

Well, you you you were like burned out. I got to get out of

Are you thinking about going back to hardware? Is software dead? What are you thinking?

Yeah. Well, I mean, two things can be true at once. I think hardware is really interesting.

Um, I'm spending a lot of time thinking about physical AI.

Sure.

Uh, but at the same time, I I've been investing for the last 10 years in application software. Uh, so I want to remain true to that. uh and this is kind of the things I've been spending time on recently.

Yeah. Take me through your current thesis on how software is changing in the age of AI in the age of you know uh you know maybe some sort of takeoff. AGI is here according to Sequoia Capital amongst many other people. uh the labs are recursively self-improving and it seems like you know Sam Alman made the made made the claim that like if you want to build a startup you shouldn't build a company that is in direct in the direct path while more and more products seem to be in the path of of AI and AGI. Uh how are you thinking about the changing dynamic with software as we know it?

Yeah, that that's a really important question and I think we've been listening to Sam a lot as as shareholders. Um, and recently just talking to our founders, a lot of them are wondering if they're just an iteration away from the models replacing what they're doing.

Yeah.

And it it puts us in the position of think what is coming next? What what is going to be defensible?

Yeah.

Um, and for us, we we've been making that prediction um last week that the next trillion dollar company will be a software business that masquerades as a services firm. Mhm.

Um and we shared that thesis with with um everyone online.

The the reality is if you sell uh tools today, you're really in the line of sight for the models

um and you're effectively competing with the next generation that they're they're going to launch. Whereas if you sell the work, you're actually benefiting from what the models are doing and all the billions of dollars that are going towards AI. Uh so part of that prediction was saying for every dollar that's spent on software, $6 are spent on services.

Sure.

Um until now we could only really go after the $1 because we were building tools. Now that we can actually deliver outcomes, companies should think about capturing those $6 instead.

So the I mean the obvious uh like first step is just get to outcomebased pricing. move away from a seatbased model probably. Would you agree with that as sort of like the most immediate change in business structure that needs to happen? Or if you're starting a new company, maybe don't start with seatbased pricing on day one. Or is the like seatbased pricing is dead? Maybe may maybe that's like an overblown uh characterization. How how do you think about just like the business model moving forward?

Yeah, we we think both can work. They just serve different purposes.

Okay. Um we have companies in our portfolio that started as these co-pilots right that are helping the workers and then we call them autopilots. The one who you know do everything end to end.

Many of them are transitioning from co-pilot to autopilot uh to to do what you're describing. A company that comes to mind is a business called Sierra that was started by Brett Taylor.

Um that basically helps companies with customer support which is

pretty direct bullseye use case of AI.

And in this example, they're actually charging the customers per outcome, right? Per resolved ticket.

Uh and so if you spend $20 per ticket resolved with humans, they charge you say $5 instead, right? And and that's driving ROI for the company. it's driving an outcome that's very measurable and so that's the first more obvious company that's really working doing that.

Uh frankly until recently outside of software engineering we haven't seen a lot of this sort of you know fully own uh end to end outcome driving. Yeah.

Um, but the models are really progressing quickly and for everyone who's been vi to end without any skills, frankly.

Um, and and basically our prediction is that this is going to happen to other categories. Um, and we've been mapping mapping out these categories and helping founders come up with frameworks to prioritize where to attack them and how to do it.

Yeah. Sorry to interrupt. One thing I' I'd be curious to get your take on. So uh if a company is just selling uh enterprise software that helps people accomplish a task that feels very threatened by, you know, a foundation model company that maybe makes an agent that can also do said said task and then you don't even need the software. So, if a company transitions to just selling uh selling the work, selling the service like you're saying, uh you would agree that they're still going to be under under threat from the labs who will also presumably be selling intelligence or agents that can go and do the work, but you're saying it will maybe be more closer to a fair fight. Is that is that the right way to think of it? Because I still think uh let's say if somebody's like spinning up a instead of making a legal AI tool, they make a AI native law firm and they're selling the work. I could still go to a chatbt a claw to Gemini and a future version of the model. I can just say like, "Hey, I need to do this thing. Can you help me do it?" So there's still going to be that competition, but you're saying that's a much better uh sort of competitive environment to be in than just selling like the software that enables somebody to do the task. Is that is that roughly right or do you is is there something I'm missing?

Yeah, I I'll just add one framework that I think is helpful for people to clarify what what that means. Um basically everyone is conflating the models intelligence with what human are really good at which is judgment. Uh and I think if you strip those two concepts that you really understand what the future will look like and so right now the models are becoming really good at everything intelligence related. So think of tasks that are um you know rules-based where just logic gets you to the goal right whereas the judgment is everything the humans really spike at which is more harder to define right it's instinct people call it taste um call it experience and those things are things we think the humans will remain really good at for a very long time until intelligence is able to absorb these skills into it but it's just really hard, right? Because we we give the models a lot of credit for their intelligence, but they're just trained on set amounts of data. Um, and so we actually think that this hybrid motion of coupling the intelligence from the model with the judgment from the humans is going to help you build the the next generation of companies which we call these autopilots, right? So your example in the law firm you will automate a lot of the intelligence work that's very much codified um which is what the model will do if you put the employee at the center someone who's really experienced who can talk to a jury who can you know mark a pause when uh at the at the right time those are all the things that you learn only from real world experience that a top lawyer will be able to experiment but that the models might will omit today. Does that make sense?

Yeah. I still I still I like, you know, uh may maybe I'm maybe I'm too agi, but I but I just assume that a sufficiently advanced agent would be able to uh capture and accumulate, you know, the ability to

too ag I'm uploading my brain right now. No, it's it's it's not that, but it's it's just the idea of like a you know, you put a smart person into an organization where they don't have any experience and they start doing things and they start getting feedback from other humans in the organization and over time they accumulate enough knowledge and domain expertise that they're able to excel at something

uh that even even using even using the the customer service example, you know, you can you can you've always been able to outsource customer service and you can go work with a firm that is excellent at customer service. The alternative has always been to effectively hire a really great leader for that part of the organization, have them build out a team and a structure and an approach

and execute uh

something that that would be a similar result. there's like a cost trade-off, but

it does feel like o overall there's like maybe some pressure on the B2B model or like the B2B TOC model if instead of uh to use your example of of customer support in uh there was a time when when software companies would sell phone tree software to not not to companies directly to do customer support but to customer support outsourcing. agencies that would then deploy that. And now you're seeing more companies go directly to like like the the the the firm that makes the the the the customer support experience end to end go directly to the company that needs customer support as opposed to having this like middle level. And it almost makes me feel like we're we're maybe going to rotate even a little bit further into DOC. When I think about law, there's a lot of stuff where the software would go to the lawyer, the law firm goes to the person, and if the person can go straight to the law firm or the the the legal model, you you wind up with a little bit more BTOC activity, a little bit more company directly to to the the service provider or the technology provider. Um, but but again, this feels a little bit dependent on specific market dynamics. I'm I'm interested to know uh your reaction to that, but also uh what does hold in your mind in the the near-term AGI scenario because there's been a lot of debate over how will marketplaces perform or are network effects still super sticky in a world of AGI or what what are what are the powers the seven powers? I guess I guess part of it kind of close out my thoughts like I I agree I agree with the entire thesis and like the relative short term like the one one to two years but then beyond that there's still some uncertainty

is yeah

around like I I don't think it I don't think it u

solves the the sort of uh discount that even traditional SAS companies are getting uh

sure sure yeah Julian

on on that point I think you're you're right that basically What we realized is no one really knows, right?

Uh I've been in many partner meetings recently where everyone just is astonished at how quickly things are moving.

I've been doing that for 10 years and it's just by far the fastest rate of change I've noticed. But some of my partners, you know, Alfred Lynn, Pat Grady, they've been doing that for a lot longer and they also acknowledge that everything is moving so quickly. And so what we're trying to figure out is also not what is going to change, what what are the things that will remain true no matter what. Right? Customers will always want things faster, better, cheaper, right? And and so that's kind of the thesis that we're describing. They're just categories that are require more intelligence than judgment. And so we think that in the immediate the ones that require more intelligence will will be actually really interesting for these autopilots because you can have you know instead of having 10 humans and one AI you have one human and 10 AIS right and just that that ratio is just going to shift as the the models get better and are able to absorb what we used to call judgment from the humans that are that is turning to to intelligence and and just to put like a practical framework to that you mentioned outsourcing It's it's actually the best wedge right now is just to focus on outsourcing if you want to capture the this services spend

because there's three reasons. The the first one is that outsourcing has already been agreed to a third party. So it's very easy to sell, right? Because it's already a vendor that's external to the company.

Yeah.

The second is that there's an existing budget that's assigned to it which you can peg yourself against. And and the third is that it it's actually very straightforward to swap a supplier. It's not so straightforward to reorganize your company, right? We all talk about headcount reduction, you know, in terms of ROI, but actually putting that in practice takes time, whereas swapping a vendor is very quick. So, I actually think that and that's the prediction is that the the companies that will succeed very well with this um autopilot model will start with uh where there's a lot of labor spend that's being outsourced. Today we mentioned customer support. Uh we'll mention also things like accounting, tax, um uh insurance and and so on so forth.

Is there a world where in the next five years or something like venture capitalists truly shift away from software and go back into rare metal mining or something crazy like that? Like I just keep I just keep coming back to like the like uh Daniel Gross uh now at Meta, you know, venture capitalist. Uh I think one of his first AGI trades, the first bet was like in the world of AGI is copper mispriced and it was it doubled in price over two years. And I'm just reflecting on this idea that like if the job of venture capitalist is to back software companies, that's one thing. But if it's if it's just to allocate capital um in in a way that can have uh exponential upside single investment return said uh would would it ever make sense to stop stop investing in software.

I would hate that because we obviously software but uh it I it has been funny a lot of people have been pushing like you know the models are so good we just need better harnesses.

Yeah. And to me, this is giving me flashback to early GPT days where people were like, "Chad, GPT is really powerful. You just need the right harness." You know, all these like rappers

that had explosive growth very momentarily

until and and so I do

most AGI pilled person ever over here.

I love it. But but but what what do you think about uh about like uh you know the the the more physical aspects of of the job like the rare metal mining for example? Does that does that seem like something people will be working on a couple years?

Yeah, I mean every everyone's talking about the fact that energy will be you know the biggest bottleneck, right, to train these models. So it it's very natural that you would go there. Personally that's not really where my skill set is. So I I really hope you're wrong in that we'll find other areas in between.

I think the the real path will be somewhere in the middle

hopefully. Right. Um and my my my way of doing things has always been to focus on talent. And so right now I'm just always asking the smartest people I know.

Um who do you know who's the smartest person? What are they working on?

And you're you're right, more and more people are going more into the physical world. Um but right now that probably means things like robotics and um things like that. Um people are making a lot of money trading commodities, but that that's uh that's a different world.

Yeah. Uh who knows, maybe that's the job of the venture capitalist in 2030, just trading copper all day long. Um,

uh, you mentioned, uh, using AI to build a website despite, uh, maybe skills that have languished since your time in software or hardware, but, uh, uh, how are you using AI tools these days? Is are you still doing mostly like knowledge retrieval, market research for your job in venture capital, or have you actually built tools that have accelerated certain things? Have you automated any of your workflows? I'm just kind of curious how people are using AI these days. Yeah. Well, I I I don't know. I hope my competitors are not listening. So, uh

they are. So, you can say I haven't touched it actually. It's useless.

Yeah. You should uninstall organic organic capital.

Pen and paper actually is underrated. Maybe you should.

Exactly. Back to the basics. Pen and paper and don't try cloud code. It's it's really really drag. Uh no jokes aside, it's been really a transformation and particularly the first quarter of this year.

Um I I mean I use it for all kinds, you know, the the work is basically sourcing, picking, winning, company building and harvesting, right? And I really have workflows that fit all these boxes.

Uh so it goes from things as you know at the end of the week it will just like scan my entire calendar, look at all the meetings I took, look at the granola notes, which is a transcript.

Sure. um do an analysis and and help me, you know, try and find the key questions and the due diligence.

Yeah.

Um and you know, it'll help me just prioritize, right? Because this job is just so much about prioritizing. It's a power law, right? There's only three meetings in the year that will really that could change your the course of your career. Um and and so making sure that you're attuned in those meetings is really important.

Uh so so that's really what I would say Claude has helped me do right now. uh and I'm sure you know OpenAI will release something next month that will be amazing and so on and you just have to experiment right and I think I think that's also like what we've seen with things like open claw being so top of mind is like this culture of experimentation is not just in software engineering now but it's it's transpired across the entire uh knowledge workforce um and you know venture capitalists are not an exception

I love it well thank you so much for taking the time to come chat with us to meet you have a great rest of your Guys, we'll talk soon.

Goodbye. Great.

Let me tell you about 11 Labs. Build intelligent real-time conversational agents. Reimagine human technology interaction with 11 Labs. You know, I if I was a venture capitalist, open claw. Every morning at 5:00 a.m., text me the name and market cap of the company that I passed on 3 years ago. I need inspiration to grind harder. That's the secret to success. As is Lambda. Lambda