Author & BRV Fund Manager: Ryan Ries

Ryan Ries: Thanks for taking the time to do this, Josh. I appreciate it. 

Josh Schoen: Thanks for having me, Ryan.

Ryan Ries: Let’s start off with your journey into VC.  How did it start, and how did you get to where you are?

Josh Schoen: What I’d say with venture is, there’s generally no straightforward paths.  I started my career working in project management at a polling data startup.  There I was responsible for implementing new features, running projects for different organizations, and building out the platform. 

As the business scaled, I ended up leaving that role.  I moved to Russia where I worked at what was the largest private bank in Russia at the time.  My background was in Russian studies, and I really enjoyed spending my time there and focusing on the region.

It was a good experience, but after a certain point I realized that this wasn’t for me in the long term, and that I really wanted to get back to the US.  I decided to go to law school.

My real passion during my time in Russia was spending time with lawyers in the bank and seeing how you could speed up legal processes by utilizing technology.  The way I thought of it was that law is sort of like the code for how society operates.  It forces people to react and respond in different ways. Because of that, people pay money for it, and it’s immensely powerful. It’s a multi trillion-dollar global industry. I found all those things very fascinating, and particularly the way that technology could disrupt and change the law.

I learned a lot in that process and during law school. I ended up working at a place called Evesort which did contract life cycle management.  It was started by Jerry Yang from Harvard, who was one of the first to use transformer-based models to recognize contract provisions and use this in an applied setting.  It grew very rapidly, and they made it to Series B round led by General Atlantic.  

Having learned so much on this journey, I ended up starting my own company which I sold to a strategic in early 2022.

Since then, I found this great opportunity with an emerging manager at Global Millennial Capital, which is a 20-million-dollar venture fund based out of Dubai.  It’s a mostly generalist fund. Where I focus is at the intersection of applied artificial intelligence at the application layer and the professional services world.  This is the world I come from and where I understand the enterprise sales process and how to build teams and systems around it.

Ryan Ries: Great, thanks for that overview of your background.

Can you talk about how these experiences have affected the thesis you have for VC.  For example, regarding the intersection of AI and legal services: what are some areas that you see as opportunities for growth, and can you juxtapose that with areas that are perhaps a bit overhyped, where AI may not live up to its promise?

Josh Schoen: Part of this is about the research into the application of LLMs. They’re essentially neural nets on steroids that are trained on a very large corpus of knowledge.

Imagine an LLM like somebody doing a job interview who comes in knowing all the questions and answers relevant for a particular position.  They would have a tight answer, know how to respond, go through all the steps, and give you that information in a succinct way.  That’s basically what LLMs are capable of, and they have the entire universe of the Internet to pull that information from.  It’s the simulacra of intelligence versus actual logical decision-making in a stepwise fashion, which is real intelligence. 

What does that actually mean at the application layer? It means that LLMs are very good at tasks that require finding information within structured text, or from generating structured summaries of unstructured text.  If you vectorize a piece of text, it’s very good at doing that.  So, key areas of growth are, for example, helping craft sales copy, things like that.

There’s also the emergence of RAG-based models, for example, in legal research drafting, memo writing, and being able to parse information from your company’s existing knowledge base and being able to build something like a Wiki from it, contextualizing and building information from that data.

Of course, no one can say what the future is going to be, but the thesis is that the new AI-powered companies in services will be able to surface, recognize, and parse information, and to innovate the new ways that people will be working with this information and building platforms around this.

To give an example, I invested in a company called Join, which is doing this in the productivity space.  They aim to simplify the process of tracking tasks and management across an organization by connecting to internal systems, being able to summarize, track, and compartmentalize what steps have been done, and what work still needs to be done.

Another company I invested in is called ScaleStack.  They’re using LLMs to connect disparate data sources and normalize data to improve sales lead scoring and to ensure that sales teams get the best information in front of them to close the sale.  They won contracts with MongoDB and Typeform and at a $4-5M valuation, that was a great deal.

Thematically, those are a couple of examples we’re seeing at the application layer of AI and how people are building the models that will be influential in the future.

Ryan Ries: Great.  And from my own use, I agree that LLMs are very good at summarizing information and giving structure to unstructured information.

We see a lot of companies that are that are coming online claiming that AI will revolutionize an industry, but as you pointed out, the key weakness of LLMs is performing a logical process that doesn’t have a firm precedent within the corpus that that it’s been trained on. Are there any areas where you think the claims of the AI revolution are overexuberant, and where it may not live up to the alleged promise?

Josh Schoen: Well, I don’t think it’s going to replace every one’s job.  Anything that is on the higher end of requiring a logical, stepwise thinking process is not easily replaced with AI.  For example, engineering is not going to be replaced with LLMs.  In my own experience coding projects using LLMs, it’s not very good.  If you start from a complete beginning, and there’s no precedent, it doesn’t do well.  If you break it into chunks, it does better, or if you want to do something like building out a simple program to test an API.  Like I said earlier, it’s good at doing things that people know how to do and have done already.  So, if there’s a bug in your code, it can do a good job of prompting the user on how to address the particular issue.  But it’s not going to fully remove the human from the loop.

What it will do is make basic programmers much better and faster.  Even 5 or 10 years ago, great programmers could be much more productive than junior developers.  But now even beginning programmers can be productive and that is in large part due to the rise of LLMs and their ability to speed up the processes of debugging and writing simple pieces of code.

Ryan Ries: The takeaway then is that LLMs are not so great at de novo creativity or operating from first principles.  But they are good at reorganizing or rehashing things that have been done before.  Those are the areas that can be disrupted.

Josh Schoen: Yes, if you systematize the process and you can explain it in steps and break it down into its components.  You can run it as an agent and do that job and keep it running.  In a sense, they can act as virtual factory workers that can keep doing a job.

But there are limitations.  As an example, an issue I know based on my legal background is that a lot of people invest in startup companies in a SAFE round.  And a SAFE is mostly considered equity, but it’s not really, at least from the traditional IRS perspective, to qualify for small business stock where the first 10 million dollars you would get 100% tax benefit on any capital gains after the holding period of 5 years.  You would have to hold traditionally from when that stock was issued.

There’s really no test or precedent to know if a SAFE would pass muster with the IRS, and no LLM could tell you.  Any LLM that would tell you that you could consider it that way would be lying to you.  It’s just not known.  And this is an example of a step of logical decision-making and gamesmanship between two different parties that LLMs cannot do.  LLMs are really like incredible parrots, not gifted deterministic thinkers.

Ryan Ries: That’s a great insight.  Getting back to VC, can you touch on what you enjoy VC, and why you like to work in the space as a professional?

Josh Schoen: The main thing is that I love working with entrepreneurs. I love being able to help them.  My brain works in a way in which I have a lot of different things that I like to know and learn. I generally enjoy helping people do amazing things, and I love seeing entrepreneurs do incredible things that frankly, I can’t do.  Things that I can’t do that they can.  And they’re laying it all out on the line and executing.  I find a lot of happiness being part of that story to help them grow amazing companies and help shift the way people work, think, and live.

Ryan Ries: I agree, and venture has a similar appeal to me.

You’ve been a founder in the past.  I think it’s incredibly helpful for any VC to have gone through their own entrepreneurial journey.  Maybe you can compare the skill set that you need to be a founder versus being an investor in venture, how they are similar and how they are different.

Josh Schoen: Yes, so in general, I would say that not all founders are good investors and not all investors are good founders.

As a founder, and because most of the time I’m within applied B2B technologies, people have very specific domain experiences in their industries, and my general preference is always for founders that aren’t actually taking a big risk in what they’re doing.  Their job is well paid, it’s cushy, and they know what they’re doing.  But the opportunity that they see by building a company is so ready, so available, and so large, that they don’t see a risk in giving up that secure job by doing it.  I think those are some of the best people.  Not every opportunity is like that, but that’s something that I tend to look for in people.

Also, for founders and VCs, both have to have resilience because both sides of the table are constantly fundraising, going from one to the next and dealing with constant rejection as part of the process.  The difference is that a founder is making an even more significant bet on themselves.  They lack diversification.  As a VC, you’re doing it across multiple companies, and you should know the math in terms of portfolio construction.  You should know how to design it and be a sophisticated thinker, investor, and being open to making bets that you don’t have to think about all the time in the same vein as an entrepreneur.

The final thing I would say that’s different in venture is that you’re there to make meetings happen and to facilitate connections, to be a bridge between a lot of different people.  Whereas as a founder, you really have to be focused on your business.  That’s your priority.  That’s what you want to do.  And for founders, going to events, doing all these other things, meeting all these extra people, it is kind of a waste of time for a lot of founders to focus too much on that.  I wouldn’t invest in those types of founders.

Ryan Ries: A good thing to think about.

What do you think about the perception of VC that popular culture gets wrong, if you were to ask the average person off the street about?  What do you think are some of the tropes that people hold on to but don’t necessarily pass scrutiny?

Josh Schoen: The biggest one is that there’s a lot of money in it.  VC blew up in a certain way that it probably shouldn’t have, since it’s kind of a cottage industry, and it’s a get rich slow, not a get rich fast kind of business.  It will take 10 or 15 years as a venture investor before you’ll see meaningful returns hit your bank account because of the companies that you backed.

The other thing is that most VCs aren’t necessarily good investors.  75% of them fail to return money to their LPs.  The people that are good at it are a relatively small group.  Also, the universe of VC of small and it’s a very specific world of businesses and industries.  And frankly, there are a lot easier ways to make large amounts of money if that is your goal.  Whether that’s in real estate, private equity, or private credit, there’s generally more money there than there is in venture.

Ryan Ries: That’s a great insight. Any parting words of wisdom for people who might read this and be considering getting into venture professionally?  As you said, it’s sort of a cottage industry, and there are very few super big firms that put out traditional job applications.  If people want to get involved, what is the best way?

Josh Schoen: Honestly, you really have to hustle if you want to do this. I don’t really recommend it for most people.  You’ll make more money working at a big tech company and you’ll need the same credentials to be able to do it.  If you’re junior talent and you haven’t worked in an operational capacity beforehand, it’ll be hard to find another job and switch, which you’ll eventually have to do.

Also, you have to have a niche, something that makes you unique.  That can be your network or the kind of research that you’re able to do that very few other people can.  There are very few people in VC with the expertise to meaningfully evaluate deep technology or biological, medical technologies, including myself.  But these companies are often raising millions of dollars based on non-expert understanding of the industry.  And that’s why VCs need technical experts if they want to operate in those spaces.

Ryan Ries: I agree with that sentiment.  A lot of companies can raise on the idea that ‘this is a cool technology, invest in us,’ but we have to slow down and ask what the practical application of this technology is. That’s where technical understanding comes in. I think this may also apply to your background, too, where a lot of people don’t understand the nuances of the law and can potentially make poor assumptions without realizing it.  Generally, it seems like as startups become more tech-heavy, that more technical experts will come into the VC space.

Josh Schoen: Yeah, that’s fair. I think in some ways, my experiences in applied data and applied AI are more useful than anything else.  Broadly unstructured documentation exists across all industries. And that’s a unique value to understand how these models are working. And I can build these things myself. Most people can’t.

The last thing I would say regarding people breaking into the VC is to learn the math, learn portfolio theory and construction, really understand how to do it.  If you have the ability to start running syndicates or pulling deals together and connecting with other entrepreneurs, then that’s what you need to do.

I think the biggest thing to do is if you’re very early in your career, do banking or consulting, and then work at a start up for a couple of years.  That’s a fairly traditional path, but there’s no straightforward path. Everyone’s got to hustle.

Ryan Ries: That’s a great note to end it on.  Thanks for your time, Josh.

Josh Schoen: Great talking with you.