VC Update:
Investing in Early Stage Enterprise AI

Ed Sim: AI is a technology that we will infuse in all enterprise software, where it makes sense and where people will pay for it. Like you, Michael, we've been around the block. Do we have mobile companies anymore? Do we have Internet companies anymore? Do we have AI companies?

Every one of our existing portfolio companies is leveraging AI in some way, some more so than others. Every new founder that we meet with an idea typically will not even bring up AI. It's just assumed. What problem are you solving? How are you doing it ten times better? And perhaps why is the new GenAI going to help you do things even better, faster and cheaper for your customers?

Michael Krigsman: Welcome to CXOTalk, where we discuss leadership, enterprise AI, and the digital economy. I'm Michael Krigsman, and today on episode 847, we're exploring how VC investors look at Enterprise AI. Our guest is Ed Sim, founder and general partner of boldstart ventures. This is Ed's second time on CXOTalk. Ed, welcome back to CXOTalk.

Ed Sim: Hey, Michael. Thank you. It feels like a time warp. I last checked, it was March 2017 when we talked about AI investing. Can you believe that?

Michael Krigsman: Ed, tell us about boldstart and tell us about your work.

Ed Sim: I've been investing in enterprise technology, basically as the first investor and partner of founders since 1996. So, I've seen a few cycles, maybe more than a few, and boldstart ventures was started in 2010 with an express mission to partner with technical founders at inception.

Michael Krigsman: We are in this AI bubble, right? I mean, there are billions and billions of dollars that's being invested right now.

Ed Sim: Today, someone wrote that we lost $2 trillion of value in the market, the worst day in the Nasdaq since COVID came around. So, yeah, I think that if you look at history, all the things we did in 2018, that was the last time or 2017. I was on your show.

All the companies we invested in, all that infrastructure, all the data labeling, all the classification, all that stuff that we built, set the framework for today. If we hadn't invested that back in the day - and back in the day, it was machine learning, not AI. It was called AI. It was categorization, those ontologies. There's a lot of humans involved.

Yes, there was some deep learning, but it wasn't transformer architecture that set us up for today. And so, if you look at the arc of 20 years, we're going to have a series of mini bubbles, and we are in an absolute mini bubble now.

And the mini bubble that Wall Street is telling us is like, holy crap. When Amy Hood, the CFO of Microsoft says I'm looking at payback over 15 years, we're just getting started. People are getting a little skittish. On the flip side, look at GitHub. I think the CEO of GitHub said that they're on a $2 billion revenue run rate and of that growth from last year, 40% came from AI copilots.

So there's a mismatch. But the point is that there's some real problems being solved with AI. And in the long run, if you're a startup founder, you're an enterprise, you have to just find the right pockets because not everything is going to be GenAI'd and not every human is going to be replaced as a VC.

Michael Krigsman: How does this affect how you invest? And as you say, we have this bizarre situation where billions and billions are being invested and yet today we have this drop in the market. So, it's like this up-and-down seesaw right now.

Ed Sim: Do the math. The billions and billions, most of that money being invested is being invested in infrastructure. It's being invested from a venture perspective in the core foundational models. You're talking about the OpenAIs, the Anthropics, and all the companies trying to build these general-purpose models.

And frankly, I think there's going to be a winner or two and there'll probably be a lot more losers than there are winners. But in the long run the person that wins will make much more capital than they lose, number one.

Number two is that's not a game for us at boldstart and not a game for a lot of VCs because it's a very capital intensive game because in my view, all roads eventually lead to the big three. Eventually every one of those large foundational model providers with all the billions being invested, and by the way, those billions being invested, a lot of that's corporate money. It's Microsoft exchanging it for Azure cloud credits, it's Amazon, it's Google and Meta putting lots of money to work.

So, if you look at a venture perspective, I go back to the core principles. My team finds technical founders who are solving some type of lifelong problem that they've really had, and they can't stop thinking about it because that's what technical founders do. They think about how to automate these problems, and then they dive in and understand what the core problem I'm solving for that user is. How am I doing it? Ten times better or 100 times better with my product.

And, of course, literally no founder we meet does not mention GenAI. There's just literally, if you'd be hiding under a rock and being time-warped five years from the past if you're not bringing GenAI up. But the best founders don't come to me and say, I have the AI of X or the gen AI of Y or the gen AI of Z. So that's part one.

Part two when we invest is I want to understand the great founders, and the problem they're solving. How big can this be? But also, I think about, hey, what are some moats? Like, for example, do you have proprietary data or can you get proprietary data that can make this that much, this model that much more powerful and that much better? How do you think about users? How do you think about the user interface? How does that make that much easier?

Look at Superhuman, for example. They're leveraging OpenAI right now, but they're just having a new growth curve. They're already growing nicely because they incorporate AI into the user experience to make it really easy.

And the third thing I would say is, when you're looking at a kind of AI over time, you have to ask the question, if I invest in this company, will they get better as the models get better? Or am I worried about being roadkill? The reason I bring that up is because traditional infrastructure investing would be like dev tools into AI, making things easier to deploy and whatnot.

I frankly think a lot of those types of companies, and we've seen hundreds and hundreds of them, a lot of that gets subsumed in the foundational model architecture. Amazon, Bedrock, Microsoft, OpenAI, they're all making it easier and easier for people to deploy. So the dev tools area is going to be a tougher area. The cool area, I think, is on the application layer. That is exciting to us right now.

Michael Krigsman: This implies that the nature of the problems that entrepreneurs are trying to solve today involve data and involve GenAI, even if they're not a, quote, GenAI company. So is the nature of the problem changing that entrepreneurs are trying to solve that fit into this mold?

Ed Sim: I've been doing this enterprise software stuff for 28 years. One is you're going to help enterprises build things faster and cheaper, you're going to help them build it more securely, you're going to help them keep the lights up 24/7 or you can build applications to help their employees become more productive, to talk to their customers in a better way, in a more engaging way. Um, and that's it. Those are just the fundamental.

So we're seeing either. I think the real question that you have to ask is, are we seeing net new AI-native companies being built that couldn't have existed before? I'll say yes, and I'll give you two examples. Or do we see an opportunity to reinvent what's already out there with a ten-fold better way to do it? I'll also say yes to that latter category.

On the former question, I'll give you an example. So, Guy Podjarny, who we've backed now a third time, he's a co-founder and CEO of Snyk previously, which is a $7 billion-plus dev-friendly security company, created a new company called Tesl, T-E-S-S-L.IO. And right now he's not divulging a lot other than he's saying that I'm creating the future of AI native software development.

And the question you ask is, what does that mean? What he's really saying is that all of the AI is being used today is looking at the existing workflow and making it better. Right? If you're coding, I'm making the coding faster, more efficient. I'm making deployment faster, more efficient. But he's saying, hey, in this new paradigm of AI development, what if I change the whole thing and turn it upside down? And so that would be one example.

Another example would be Protect AI. Protect AI just announced a $60 million Series B yesterday at a $460 million valuation. And they are the leading end-to-end AI security provider. And by the way, we started collaborating with the founders in October of 2021, a year before ChatGPT even came out. So that's how early we go. We have to envision new markets. And these founders envisioned the new market. So that would be kind of net new. And I think these are going to be massive, massive categories, massive ways to do things.

And then reinventing the old would be like companies like a Superhuman. Superhuman is kind of that company that's reinventing email. They've been around for a long time, but now they found this ten x growth curve with how they're deploying AI into email. So they're sitting on all your emails, they can help you create personalized emails, they can summarize the email, they can write emails for you, and it all does it in a super-efficient way, easy, elegant interface that people love. So those are some examples of how we think about it.

Michael Krigsman: Please subscribe to our newsletter and subscribe to our YouTube channel. We have amazing shows coming up. What about traditional business process software? Is that getting sort of pushed to the side? Or are founders rethinking those types of business processes through this lens of AI and data?

Ed Sim: They are absolutely rethinking it. And by the way, I do think that some of the best application layer companies, in order to be sticky, they not only have to provide an answer, but they have to be part of a workflow, or creating workflow. And that is the point that you're asking.

And then I ask the question to a lot of people, like, okay, let's take a step forward. If all things go right, how much of this end-to-end workflow, which touches ten different systems and may require ten different tasks to solve with the human in the loop in the middle, how much of that can you automate and how much will you not?

And frankly, we do have a hallucination problem in the enterprises. You have to be very careful about which areas you're exposing the customers to say, hey, I'm 100% confident that this is a great answer, versus, uh, answers that may screw things up. I think there was an example in the past where someone was chatting with an airline interface, and apparently, they got a $10,000 credit or something ridiculous like that, right? So, you have to watch for those things.

But we have a company, by the way, called Crew AI, crewai.io, which is an open source platform that allows anyone to build their own agentic workflow, workflows in a super easy way. And now we're doing, we're running more crews or agents. They're doing a million runs a month right now already and growing super, super fast. We're watching the enterprises adopt these super quickly.

And I think that the old RPA—you and I have talked about RPA in the past—robotic process automation—was fascinating. UiPath was a killer company. Automation Anywhere was a killer company. You know, we were in Fortress IQ. We sold that to Automation Anywhere. Catalytic, Sean's company, was sold to PagerDuty, but we've got the next version now.

And I truly do believe in the enterprise. You will see not just point products of answering a question, but workflows being built. There's going to be vertical silos where people focus on certain areas like compliance and really go very, very deep, where maybe you might use a traditional foundational model like an OpenAI to ask an initial question. Then you may shovel it into an open source model sitting on your own servers to make sure. Then you train on your own custom data to keep it private. And then there might be a human in the middle answering that question.

The ultimate end goal is going to be increasing efficiency, I think by two, three, four, x versus replacing everyone. I don't think that vision is really real yet in the enterprise in order.

Michael Krigsman: To drive these tremendous efficiency improvements in every case, or almost every case, is the problem being solved again through that lens of the data?

Ed Sim: I have never seen in my 28 years of doing venture investing and enterprise technology just the enterprises just climbing onto the idea of AI, where it's a board-level mandate, where every CEO is talking about in every earnings call, where every CEO is going down to their CIO and CTO and saying, what are we doing in AI? How are we doing?

Right now, what we're really doing is just scratching the surface. Even you're looking at all the Morgan Stanley and Goldman Sachs CIO surveys. I'll say, hey, 20% of the stuff is in pilot; most of it's not deployed yet. Maybe only 5% is deployed in the enterprise. This is going to take a few years to roll out, three or five years, but people are investing in it.

By the way, I also think that the data, the proprietary data, we're only getting started in that right now. Right now, the low-hanging fruit is just using these foundational models and having their employees be a little more productive, maybe uploading some documents and getting an answer. But as you move into more sophisticated workflows, as people take models and do it on-prem or maybe with Azure enterprise, you can actually really make sure that your data is safe and private.

We're just scratching the surface now. That's why I'm saying that we're in a mini-bubble. But in the long run, if you keep staying the course and choosing wisely, you're going to have tremendous opportunity moving ahead.

Michael Krigsman: We have a really interesting question on Twitter from Arsalan Khan. And Arsalan Khan says this. He says most enterprise software is bulky and archaic. The data within this enterprise software is locked in. For AI to really be implementable, don't we need legacy enterprise software vendors to play nice? But why would they if it hurts their bottom line?

Ed Sim: I really think it depends on what the question is, quote-unquote legacy data, because ultimately the data all sits in a database or may sit in a data warehouse once it's moved to a Snowflake or a Databricks or what have you. Both of those companies have an easy ability to extract insights and build applications on top of it.

I think the whole idea of playing nice, sure, there are going to be certain areas where people get locked in, but I think over time is that if the data is all in one place because it is. That's the biggest thing. When I talked about from seven years ago, the whole idea of data warehouses and data lakes. Now you can store infinite amounts of information, whether it be unstructured data or structured data, in one place. It's making it easier and easier.

That's why Databricks is a great company. It's making it easier and easier for people to build applications on top. So I think unless you're talking about old, super old stuff where it's hard to get that data, it's available right now. And that's what people invested in seven or eight years ago. And that's basically what I said before, was that if we didn't have that initial era and the initial hype and bubble, then we wouldn't be where we are today.

Michael Krigsman: And so, a big part of what founders are doing then is figuring out how to take advantage of this large volume, huge volume of collected data in order to make more efficient or drive innovation, whatever it might be.

Ed Sim: I'll give you a great example. We just funded and just my partner, Shomik Ghosh, just did a deal on a company called Noded N-O-D-E-D. So, and the founders, one guy was Steve and Chris. One was head of architecture and infrastructure and integrations at Slack. Before that, he had sold some integration companies into Salesforce and Dell. The other founder, Chris, was the COO of Dell Boomi.

And they're rethinking how employees interacted with applications using AI. They have a traditional note-taking interface. As you're typing, it's integrated already with many of the applications you use. And surfacing insights popped up instead of opening up six windows. If I'm typing Michael and et cetera, et cetera, like, hey, Michael, I saw that Joey on the team sees you in Salesforce. Here's the data in Salesforce, or here's the data sitting in Gainsight, if you take a look at N-O-D-E-D.

So you can see a new way of folks saying, how do I create this integration layer, but from an end user first? So there's tons and tons of opportunities where to your point, there could be data siloed. But how do we unlock that value? How do we make it very easy for the end user to not leave the context and bring the context to them?

Michael Krigsman: And as you say, the way the data is stored today, in contrast to, say, seven years ago, it's much more accessible. And that trend obviously is going to be continuing because.

Ed Sim: Yeah, and the tools are that much better, too. And in fact, if you think about it, the tools, what will GPT-5 look like? The tools will get better and better and better. See, once again, it goes back to that prior point is maybe, perhaps you will be building something. And you know that I'm ready to swap out any model at any point in time.

Or, if you think about it, in any enterprise workflow, there might be ten steps in a process, and maybe you might have to ping, as I said, a general-purpose foundational model. Maybe you ping your open-source model; maybe you ping another model, or maybe you have a RAG vector database sitting next to it to pull proprietary data. You might, in that workflow, have three different models that are being touched for very specific reasons. And that's some of the architecture decisions people make now as they build out these workflows.

Michael Krigsman: We have another question from Twitter, which is from Mark Loudvenust, the number three. And he says, what do you, Ed, think about using AI to automate workflows in the mortgage and banking industry?

Ed Sim: I'll zoom out and just say a lot of the largest financial services companies in the world are already building their own applications internally to automate some of this. And if you think about kind of that workflow in the mortgage and banking industry, there are a couple of things you still need to think about.

I know that even before this gen AI wave folks were using, we'll call it machine learning versus AI, even to make credit decisions. And I don't know if you remember the Apple fiasco where credit decisions were being made in a wrong way. So I think that these things are happening. But they're also regulated industries where banks and mortgage folks have to be very, very careful about how much they trust the AI, or at least making sure that the AI is trained on very specific data which will not provide any bias. And that was a very, very big deal.

But I can tell you this, I talked to a lot of the banks, I talked to a lot of CXOs, so we're on CXOTalk, which is great, but they're all building something right now. But that is one thing they're definitely deathly afraid of, because if you do it wrong and there's a backlash like we saw with Apple, Goldman Sachs go look it up. You've got to be very, very careful.

But I see areas more like KYC and fraud. Those are the easier areas to go. If you think about KYC and fraud, well, particularly fraud, that's more of an internal thing. The interface is more internal. It's helping them find the needle in the haystack. And then of course they'll reach out and say, proactively cut you off your card. The stuff that is more customer-facing when it comes to decision-making and the higher value and the more at stake it is, the longer it will take for those applications to be fully deployed.

Michael Krigsman: On CXOTalk, we've had a number of folks from financial services, so go on our site; we've had the chief data scientist of Dun and Bradstreet, the chief analytics officer of FICO, who's going to be coming up as a guest. We don't have him scheduled yet, but we're looking at the schedule now and a bunch of others. The chief data officer of Goldman Sachs. So, just search for these on cxotalk.com, and look there because it's a great place to begin.

Ed Sim: That's a great group of people. And look, they're all working on little hanging fruit. I remember seven years ago, the ability to read unstructured texts or read a PDF and extract data was not really good. It was OCR technology. Now with GenAI it's much better. I can pull up any picture and pull up the data and be 99% accurate.

So those things are all getting incorporated to the workflows. But what I'm talking about, if there's an end-to-end solution, that will take a while.

Michael Krigsman: Ed, you talk about inception investing relating to the very early stage. Tell us about that.

Ed Sim: We like to be kind of that partner with founders, helping them collaborate on their ideas from the very beginning. And what happened when we started boldstart in 2010 was that the earliest you could go was seed. So that category was created and then suddenly there's a new category created five years later called pre-seed and pre-seed washing. Presupposing that you would need a seed round after that. So then there's a new category called seed plus.

And when you look at the data, the point is who cares about all this terminology? So I and my team came up with the idea of inception investing. What that is just what we've always done is when you're a founder, technical founder, with an idea, who do I go to? The point is that whether I want to raise 500k or whether I want to raise $20 million, I should probably be going to the same place because it's the same idea to get started, to hire your first eight people to have that search for product market fit.

It doesn't matter how much you raise because you can't force yourself to. So that's why we created this idea of inception investing. And I think it's starting to take hold now. Right, in the sense of who do you go to founder? Like if I did a pre-seed round, do I go to a pre-seed fund? Or if you're a third time founder and you're raising $20 million? I'm not raising a pre-seed round. Is that an A round? I don't know. Go to an inception investor. If you're starting and you want a check and you want a lead, who's going to partner with you from the very beginning? Who is okay. Before you even write your first lines of code, then that's who we are.

Michael Krigsman: How is this different from traditional angel investing?

Ed Sim: Angel investing in traditional parlance is that it would probably be the same stage, but angel investors are usually you're thinking about these smaller angels writing 25 to 50k checks. So it would be the same stage as just having that idea or going and do your friends and family around, but you're actually having an institutional investor partnering with you who can help you accelerate that path to product market fit.

I'll give you another good example. Let's just say you're two founders and you have three more people you want to bring onto the table. You can either cobble together 25 50k checks from 20 different people, or you can come to one person for one to two or $10 million, and we can even help you pre-sell those initial hires to reduce the risk of joining this business. To make sure that you're not going to have to start a company, raise an angel round, and then continuously raise capital for the next twelve months.

Do you know how distracting that is? Why not get it all done right at once? We can get started and you have your six or eight people and you don't have to raise any capital for a long time. All you have to do is build a product, get the product market fit. And I think that is a much more efficient way of doing things.

Michael Krigsman: So, you're like a one-stop shop, so to speak, for very early stage.

Ed Sim: Investors, for founders, yeah. So come to us. You have an idea, are technical, and're building something in enterprise. You want to leverage our CXO network. We're the folks that do it. I mean, look, we have a history of doing this. We have companies like Snyk, which is worth over 7 billion, BigID worth over a billion. Security Scorecard worth over a billion. These are all companies we partnered with at that stage, and then we have others like Superhuman or Customer.io, which we sold for a fair amount as well. So, we know how to do this. We love doing this. And there's a special kind of patience as well. When you actually work and invest at.

Michael Krigsman: At this stage, how do you make those decisions at an early stage where the founder may not even have a business plan yet? It's an incredible risk for you.

Ed Sim: I don't think it's incredible risk. It's the way that venture used to be done back in the seventies when Arthur Rock started it. He would work with people, help them build their plan and fund it when there's technical risk. And for us, it comes down to a few things. One is it's the founder or the co-founder of that team. What gives them the right actually to build what they're building? To provide the unique insight they're building. That's ten times better than what their customers are using today. And what gives them the right to have the insight?

To see the future, we need founders that can see around corners, that can see the future, and usually in the enterprise space, Michael, it's folks that have been living and experiencing a problem over and over again. When Alex started Security Scorecard, he was a former CISO, and he kept asking himself, what happens when I have this whole vendor risk assessment? How do I track all the security issues with all of my vendors? And how do I do it in an automated way?

That's how the Security Scorecard was born. That is the path of a very great. Versus someone coming out of business school that has zero experience and saying, hey, I'm going to go after building a next-gen security company. That's pretty rare to be able to do something like that if you have not come from the industry.

Michael Krigsman: So, Ed, can you describe the characteristics of a founder who or the situation a founder, maybe a couple of them come to you. They don't have a business plan yet, but they have what they think is a great idea. And I'm sure you see this all the time. What's the difference between the ones where you actually write the check versus the ones who you give a pat on the back as they go out the door?

Ed Sim: Founders have to have the ability to zoom in and zoom out. They need the ability to zoom in, to get into the mock and show me that they know how to build a product and they understand the problem that they're solving and for who. And they also need to be able to zoom out and say, hey, if everything goes right three to five years from now, how big can it be? Let's be honest, not every founder should go out and raise venture capital.

If you decide to raise venture capital, then you're probably signing up to try to build something big to make a really, really big debt. There can be lots of great businesses that get the five or $10 million of revenue, and it's more of a lifestyle business where you don't need the funding. So, I think that's part one.

I think part two is that these founders typically just have passion. You could just feel it when you meet with them. It's just burning inside of them. They're obsessed about, obsessed in a good way about solving this problem. Right. And so that's what they're constantly thinking about and talking about. In fact, maybe it's been their life's mission. When Brad and Jeremy started Customer.io, that was their third customer service company that they had started, and they had a new architecture to kind of build things.

And the last part I would say is that we always also like before they write their first lines of code is to say, hey, who are you bringing with you? What is your history of success? You don't have to be a former founder. You could have been head of product or head of engineering at a startup, an employee five, where you helped build some very important product and got it launched in a very big way, and now you're ready to take that step up. That's okay. You don't have to be a former founder, but I want to see demonstrated history of success, of passion, of picking yourself off the floor after you've been beaten down a little bit and kind of persevering.

And then secondly, if you're really good, you're typically going to already know who your first five to ten people are that you're going to hire. You may not hire all ten of them, but I want to know who's going to follow you, who's going to work with you and why. Those are some very important metrics.

And the final piece, of course, is to my earlier point is if this all goes right, how big can this be? And I had this saying internally and externally, is that it's not the TAM or total addressable market size that you begin with, it's a TAM that you exit with.

So I'll give you a great example. Snyk now owns the developer-friendly security space. When they first started, when Guy first started, he was going after the idea of how do I make sure that all the open source packages that you're using are constantly updated and we find all the vulnerabilities from a security perspective, not only did he not start in open source, he also dove deeper and started in JavaScript and more particularly Node. This was as narrow of the wedge as you can get because his question for himself and for everyone was if he should go a mile deep or a mile wide.

And I'll give you an example, do I start with JavaScript and do Ruby and Python and everything else, or do I work with the JavaScript community in a particular Node and dive very deep where not only would I find the vulnerabilities, but I would also offer the fix. Now Michael, do you think a Node developer cares about whether you have Python or Ruby? No, he went down very, very deep, and developer-friendly security not only found the problem but fixed it. For me, then, the question was how can we add other markets? And now, if you look at the company now, we're actually much, much broader than we are. But that's an example of if you start really narrow, the real question is how big can you get and what would that look like?

Michael Krigsman: And having that clarity around the nature of the problem, and I'm assuming being able to communicate that really, really well.

Ed Sim: Absolutely right. And sometimes it doesn't reside in one founder, it might be two founders, one might be more the technical person and the other might be more of the go to market, you know, kind of person who's more external and sometimes like in a Guy Podjarny, it lives in, it lives in one person, right? It just depends.

So those are some things we look for on the AI side. I do like to think about kind of the idea of if you assume that, um, X is going to happen, what happens next, right? And I like to call that the second order of effects of AI, right? And so for example, if you assume that, um, 30% of enterprise developers or developers are using code pilots right now, and maybe it goes to 40%, what's the end effect of that?

Well, the end effect is that there's going to be more code written, a lot more code written. What happens when there's more code written? Well, you're going to have to secure that code. How do you actually maybe scan all that code before you run into the models and look for vulnerabilities so you don't have more secure code? Or how do you help developers keep up with all that code and do code reviews? How do you help them?

So those are all the questions you start asking is what are the second order effects? If x happens, then why? Because if you're only investing in what's happening today, then I'm going to be roadkill. I can't invest in what's happening today. I need to invest in what's happening next, and that's how we need to think about it.

Michael Krigsman: And how do you develop for yourself the confidence that this projected future that Joe or Mary is presenting is the future of the world?

Ed Sim: Some of it, I call it, is intuitive TAM and intuitive understanding. Right. I'll give you a great example. When Ian Swanson and Daryan Badar, we started collaborating with them in October of 21, Ian and Dee were like, look, I'm sitting inside of AWS. We're leading go-to-market for AI, ML, SageMaker, et cetera. We found that while the chief data officers have full visibility into how the models are performing and what models they have, security has no vision around that, and we're starting to get asked about it.

And so as we started working with them, we're like, what is the product going to be? How's this going to look? How do you get that to market? And frankly, when we invested, we're like, look, I don't know. I can't predict the future in terms of when the market's going to develop. But let's just make sure that this founding team has three years of runway, because within the next three years, we believe that there's going to be a seminal moment in AI, whether it's a massive hack or some other thing that happens, in which we're going to have to secure all of the AI and ML models and the data and the whole chain around it. That was the vision that was. We closed investment in early 2022.

Guess what? We got lucky chat. GPT came around at the end of the year, and all of a sudden people were like, hey, how do I secure this stuff? How do I scan my Python notebooks and my Jupyter notebooks for vulnerabilities? How do I provide a dashboard? Fast forward? The company's been off on a sprint. I've never seen the demand come in from large enterprises like I do right now because they're very hesitant to deploy AI without securing all their data and their privacy. So, I like to say there's no AI in the enterprise without AI security.

So that was intuitive. There wasn't a Gartner TAM or Gartner roadmap or anything else. It's like, hey, is AI being around in the market? Yeah. And if it works, sure, that could be absolutely massive. So that's kind of the faith you need to have. But the founders saw around the corner they had evidence of how this could look. And of course we talked to some of our CXOs as well, saying not a problem today, but it could be. And guess what, three years later it's an absolutely must have now problem right now.

So we're lucky in terms of timing. It could have been five years or seven years. So the question is could we have survived across the desert and made it across there? And I'd say yeah, we probably could have because they could've found their own 20 customers just to get us going through that process.

Michael Krigsman: Ed, when you invest in this way, that relationship will likely extend over years. So, to what extent is your liking that person and feeling comfortable with that person just on an emotional human level? Where does that come into play?

Ed Sim: If it does, we're investing in people and it's not only me, it's the founder. The best founders have the ability to choose who they want to work with, period. Period. Or we may find the diamond and the rough where no one wants to work with them and we're fortunate to dive in and partner with them. But look, I mean, at the end of the day, what I feel very fortunate about in our business is the ability just look at our NPS score. When founders come back, when they start their company, their new company, they could probably go anywhere else. They could go to lots of other places. There are lots of firms that have better brand names than boldstart, frankly.

But we definitely bring a human element that is so important to us. Rahul from Superhuman, we funded his first company, Rapportive, when he sold to LinkedIn. We were his very first investor in Superhuman and continue to work with him. Guy, he's on company three now. It was Blaze, sold to Akamai, then Snyk. And now there's new company Tesl, right? Brad and Jeremy, Customer.io 2.0. Well, we just did Customer.io 1.0 with him when he exited a meta. When they came back and spun out, it's us again at the table with some of our friends from Redpoint and Battery.

Over time, it's the relationship. And sometimes, by the way, not every company is going to succeed. But it's how I work with founders that actually don't work out to the goals that we had and how the founders work with us. And that might be the best learning experience we can both have to go and start company number two together.

And in the case of even Protect AI where I brought up, Ian and I have known each other for ten plus years. I tried to sell a company to him when he was running DataScience.com before he sold it to Oracle. So we got to know each other that way. And my partner Elliot knew Ian from kind of the, uh, 15 years, I think. So, these things have a very, very long lead time and long life.

And um, you know, if you, as a founder, are going to take a call at twelve midnight, or as an investor is going to take an urgent call at twelve midnight, then you better really like each other as well as you kind of go through going through this process. And like, means that, that also means that, um, you have the ability to have a relationship where you can also be tough on each other, right? Because otherwise that's not a good relationship either. That's what really good friends do, is they pound each other, but also challenge each other and make each other better. That's kind of what we feel grateful about in terms of working with founders and vice versa.

Michael Krigsman: Now would be an excellent time for you to subscribe to the CXOTalk newsletter so we can keep you updated on our upcoming shows and subscribe to our YouTube channel. So go to cxotalk.com and do that right now.

So here is a question from @textshiva on Twitter: Do you think only large companies like Google, Amazon, and Meta will successfully implement AI? Or can small-scale companies also leverage these technologies effectively? So what advice would you give to small-scale companies who are looking to adopt AI? I'm assuming that he also means small-scale startups. When he says adopt AI, he means to develop.

Ed Sim: Software, I'd say the answer is both. The large companies are going to be the, as I said, all roads eventually lead to large companies when it comes to building a general-purpose foundational model. So, but because that they're building these general purpose foundational models, and because the Meta's of the world are open sourcing all of their stuff for us to use, then as a founder, you can get the benefit from all that capex investment that they're having.

And over time, these models get better because they're investing in it, and the cost to infer and train will get much cheaper. So if you're a founder, the question that you have to ask yourself is, if I'm building something and solving a problem, first focus on the problem you're solving first. And then, can AI turbocharge or make it better? Then ask yourself, when these models keep improving, is my application going to improve? Is what I'm delivering to the customer going to improve?

And as I said, some people say incumbents win, some say startups win. Some people say, I had this saying that there's a Goldilocks type of company that can win too, which is you're five or six years old, so you're not a startup like a first round startup, but you're also not an incumbent and you have the ability to move very quickly and ship product and you also have data that you're sitting on. So there's no right answer.

All I'm telling you right now is that you just have to be super thoughtful about who your customers are and why they're going to pay for whatever you're going to deliver. And it's the greatest time in the world to be a startup. Just grab an API. Grab an API and deliver something, or grab something from Llama for free, roll it up, train it on your own stuff, it's out there. This is democratizing AI for everyone, so everyone has an opportunity. And that's why we're excited and thus.

Michael Krigsman: The role of the passion and the commitment you described earlier.

Ed Sim: Being so and understanding the problem you're solving. And why do they need this, why are they not getting it from an incumbent? What is it that you're doing that's so much special? Right? Why are you moving faster, right? Or why is this interface better? And you have to know your space.

Michael Krigsman: So, to paraphrase, it seems there are several things; you have to understand the problem as you were just describing.

You have to have the passion and the commitment to move forward. And you also need to understand the available tools and the nature of the data that's out there so that you can put these pieces together.

Ed Sim: You also need to know what your competition looks like. Is it a homegrown solution that someone's building? Or what are the incumbents doing? And why do you have an advantage? Or why do you think you have an advantage? And that goes back to, once again, typically, when people start enterprise companies, they don't come from the consumer space. They actually have lived and solved and used these tools before. I think that's a big part of it.

Michael Krigsman: We have an interesting question, an enterprise question from Arsalan Khan. And Arsalan is a regular listener. He always asks these thought provoking questions and he says this when we do business process re-engineering, oftentimes we come across business processes that are not documented and often culturally related to the organization.

To me, that's a very nice way of saying companies have these byzantine processes that make no sense at all, but they're embedded in the fabric of the organization because, well, that's how we do it. And his question is how can AI help in a case like this where the information is not even there to be surfaced?

Ed Sim: I think it depends, right? As I said, I tried to go down this path back in 2015 with a guy named Pankaj Chowdhry, who did Fortress IQ. And what he was trying to do was use machine vision back in the day, using GPUs to analyze the work being done on a desktop, to go back and reengineer and re-understand the business process that folks were building. And you fast forward to today. We are more wired and more connected today than we ever were. And so I would say that every employee has exhausts that they leave behind, cookie crumbs they leave behind of what they do every single day.

There are companies for example, I'll give you a great company in our portfolio called Grip Security. Grip Security has built a knowledge graph. And the knowledge graph starts with the person, and it actually goes to the person, the application they're using and who they're connected to and what files and data is being shipped around. So there are lots of tools like that. And by the way, they use it for security. They have a category called SaaS security posture management. So when you look for anomalies, but the point is that the technology today can sit on top and look at all the exhausts from the employees, look at who's connected to who, what files are going over where, and say, hey, I think this is a process, or I think that's a process, or that's a process.

And then you can work with the head of business transformation, say, is this the most efficient process? Is this a process? And if so, then how do we build some AI into the workflows, maybe scanning some PDFs or doing things? And so that is happening as we speak. And every company that is in the large scale enterprise space is working on building this stuff.

Michael Krigsman: Clarity on the nature of the problem and the depth of why this problem is important. And what will potential buyers care about with today's technology?

Ed Sim: Everyone's digitized the world. I mean, versus ten years ago. We're more digital than we were before, and there's more SaaS apps being used than ever before. And it's much easier to track and find and understand who's connected to what than it was 15 years ago when everything lived in silos on people's desktops, on their desktop client software. It's only getting easier and easier and easier to understand what is a process.

Michael Krigsman: We have another really interesting question from LinkedIn, and this is from Ralph VonSosen, and he says, how do you see industry-specific AI evolving? Highly specific vertical industry focus and then broaden from there, or very specific use case, like a new application for insurance or communication services, and then broaden from there.

Ed Sim: AI and vertical use cases once again gets back to the data that we're talking about, Michael. Right. So I'll give you a good example. Let's just say for compliance, what if you're looking in the chemicals industry and there are thousands of pages of manuals on how to ship hazardous materials and to make sure that people are doing it the right way? how do you understand all of that? That'd be a perfect example for vertical AI compliance software.

The question really is that a lot of the bigger companies who are more innovative are going to build themselves. And then the question is if you build something, what does that middle market look like to start with? And eventually can you build something so good that can replace the endpoint? Then of course, I think what you're really talking about is horizontal opportunities. What if you're building out a next gen customer support system? Is that something that should be vertical or not?

My answer would be, I don't do vertical investments. It's just nothing that we've done. Frankly. We've always gone for more horizontal swings of the plate. But I do know lots of folks that believe that the vertical AI for X is going to be the next big thing. And AI does lend itself well to that because of the specific nature of the data. So I just think there's opportunities for both. Frankly, there's not one way to skin the cat.

Michael Krigsman: Here's from Shiva again, who says, and there's actually somebody else that's asked this. From your perspective, what are some common mistakes that companies make when implementing AI and how can they prevent these pitfalls in the AI era?

Ed Sim: The first common mistake is there's zero differentiation for what you're building. So you're just an AI wrapper and you literally are just offering something that a lot of the big companies can offer. Right. So I think you really once again have to go back to understanding the space that you're in and the problem you're solving.

Two is let's talk about enterprise. We're talking about enterprise workflow here. People, I think, underestimated the importance of data privacy and security. So if you're going to sell into a large enterprise, and the larger they are and the more regulated they are, the more they're going to ask you about, can you deliver this on-prem or can you give me an ability to make sure that this data is not being shared to train other models, et cetera, et cetera. So I think people do underestimate that issue and just believe everything is just cloud based.

Michael Krigsman: Another question from Lisbeth Shaw, who wants to know how does this micro froth that you described earlier or the hype affect investing in AI startups today? And what does this micro froth look like?

Ed Sim: Well, the micro froth looks like today is a $2 trillion red screen across the board for AI and people saying please be patient because we're investing for the next ten years, and that's going to take the bloom off the flowers. But I think this is the opportunity for the real builders to continue building and stay the course. Don't let's not look at the vagaries of the market every single day. But if you look at five or ten years from now, is what you're building, once again, just can I build 100 or 200 or $300 million revenue business from the opportunity I'm building? And am I offering a solution that is ten times better than what's out there?

And if you believe that, who cares what the market does? Just keep building what you're doing. Stay efficient and the models are going to get better, period. And you will get better and your product will get better. So these are opportunities for us to step up and invest even more. Maybe when people are being fearful. You can't always jump up and down based on what the markets are doing or what people are saying. You've got to have a steady belief.

Michael Krigsman: Are you a coach, a bank, both, or something different?

Ed Sim: We're more of a partner and a coach than a bank. Look, if you look at venture capital, there's lots of venture capital out there, there's lots of venture capitalists out there. Money's green. But the question is, what are you going to do? What's the relationship that you have with the founder, and what's the relationship that the founder has with you and your team? And so I think at the end of the day, we think of ourselves more as partners and coaches than banks because money is a commodity.

Michael Krigsman: We have another question from Arsalan Khan, and he says big organizations are good at disciplined processes and momentum, but want to be agile, to react quickly to markets. Small organizations are agile but want that discipline. They don't have it for startups.

What's the right medium, the right balance between the discipline and the agility for startups?

Ed Sim: When we invested in inception, the early signs of success for us, number one, is product velocity. Everything that this organization does should be focused on shipping a product as quickly as possible to put it in people's hands so we can figure out how good it is and learn from it. That's number one. Number two is learning velocity. That is, how quickly can the founder create experiments, test those, quickly decide whether this is good or bad, and then adjust on the fly.

So, in the very early days, it's all about speed. If you layer on too much process at the very beginning, you're going to sacrifice that for speed. And as you get bigger and bigger, then of course there's more processes being put into place as you have customers and everything else. But right now, I can tell you this, the only balance and focus is speed. Ship product out the door the second we invest, and everything else is a distraction around that.

Michael Krigsman: Ed, to what extent does FOMO, fear of missing out, drive investment decisions? Whether you or your other VC colleagues at other firms?

Ed Sim: That's a running joke. FOMO. Look at the other day. Perhaps this market pullback today is going to get people a little jitterish and perhaps there'll be an opportunity for the, for the steady believers who know that in the long run, if you choose wisely, you can build really big businesses. And so, look, it ebbs and flows. That's why we're in a mini bubble now. Every VC needs to have their AI play. There's a very specific reason.

Also, we don't call ourselves AI investors. As I mentioned before, it's just infused in everything we do. And so most founders are leveraging it, but we're not an AI investor. So FOMO does drive the markets up and down. But if you look at it overall, it's going to be like this. Michael Right, every time you look at a tech boom. But overall, the question is, is this a platform shift? And this, in my opinion, in the 20 years of doing it, is one of the biggest transformational platform shifts I've ever, ever seen. And I truly think we're just in the first inning and those innings are going to go up and down and up and down, but in the long run, the curve is going to be up and to the right.

Michael Krigsman: As an investor, how do you decide through the vagaries of this up and down, what to stick out and tough out, and what to just simply say this is not going to work.

Ed Sim: In the early days you can see that product velocity and how quickly folks ship and at every stage it's different. But I think the first perspective would be, can you ship product? Can you get the product market fit? Can you get to a repeatable growth scale. Those are the three questions. And every step of the way, you got to work with the founders and say, are we hitting these marks? Are we hitting these marks? Are we hitting these marks? Are we able to get continuous funding from the next round? For the next round? For the next round? Hitting our milestones?

When things get challenging, because they always do, you have to look at the founder and say, hey, are you running out of energy? Or is it just taking six months longer or nine months longer? And if so, why? Right? And so there's always those nuances around that. But it comes down to the founder as well. Or by the way, the founder may say the market's not there. I was ten years ahead. The market's not there. But a lot of it is driven by the founder and our work with the founder.

Frankly, it's not just us founder as well. And you know what happens, startups die when founders run out of energy. Because I've been involved in lots of companies where founders have pivoted twice and then all of a sudden has created some amazing thing when founders run out of energy, because it's really hard and lonely being a founder. It is mentally taxing, it is stressful, it is your identity. And sometimes it can run out of energy, and sometimes they may be relieved. You say, are you okay? Is this going to keep going? And you know what, your opportunity cost may not be worth it right now. Or a founder may say, look, I know it's just around the corner. I'm about to launch the product. I've got three customers ready to go. And so really it comes down to the nuances. There's no one answer, but it's a lot of, it's founder driven.

Michael Krigsman: So you're making the distinction between the founder and their energy capability, all of that versus the market, because you can have a founder who's run out of energy, but the market is right, but they're just, you know, they're not the ones to continue, for example. Or they want to continue, but the market isn't ready, for example.

Ed Sim: Exactly. That's where it's a nuance, right? It's both. And you have the conversation, but a lot of it comes down to that relationship that you have.

Michael Krigsman: Can you just summarize common mistakes that founders say when they come to you?

Ed Sim: First thing is just be prepared, right. Kind of know what we do, how we do it. I mean, what we look for. We have lots of content out there on our website and through our social etcetera. I also have a newsletter called www.whatshotvc.com so I talk every week about kind of what I'm seeing in the market. So just do your research and homework. Know why you may be a fit for us and tell us why that's going to happen.

Two is I think you have to make sure you get your story straight really succinctly and be able to tell me in two sentences why what you're building. Tell me what you're building and why as succinctly as possible. And let's expand from there. And as you're presenting something, you always have to make sure that the next question on the slide is how do you get us to ask the question that it might be on the slide next? How do you get us to turn the corner? Turn the corner.

And the final thing is that some founders think they come in and just say, hey, we're getting a check today. But the real question is how do you think about building a relationship with each other and getting to the next meeting? That's the goal. If you get to the next meeting or get us to say let's meet again, then I think you're winning.

Michael Krigsman: What about personal charisma? How important is that?

Ed Sim: Look, at the end of the day, we have founders do lots of pitches. They're probably looking at the ten VCs they pitch in a day. And we see lots of pitches too. And I think for both founders and investors, the energy, the excitement, the passion around, the questions, the engagement, I think that's all very important. I know a founder when they jump off the screen or jump out of the room in the meeting, and the founder knows the same when the investors almost finishing their sentences.

And I think for both sides, you got to find the energy and passion because look, there may be some amazing companies. I don't wake up excited about just wanting to invest in that area, I can tell you this. I love cybersecurity. I love partnering with my friends from Israel, and we've got lots of cybersecurity companies in Israel as well and whatnot. But there could be some categories that just don't fit us personally, and that's okay. That doesn't mean you won't build a multibillion-dollar business; just doesn't fit us personally and vice versa. So founders should find that person equally excited for them and their project in the market as we are.

Michael Krigsman: So again, there's this aspect of the chemistry and the comfort and the general willingness and sense that having a relationship with this person, this VC, in your case, or startup founder, will go on for years. And this seems like a good thing.

Ed Sim: For me, whoever, and hopefully go on multiple startups, right? And I think, Michael, you just nailed it. And the reason I keep saying this is because the industry got away from this during the ZIRP era. Basically, I felt like the world moved to a transactional nature versus a relationship-driven investment where people were dropping term sheets within a day or two and just doing background work and not getting to know the people. And then what happened was when the shit hit the fan, and you had to actually make some hard decisions, the investors and the founders didn't know each other so it's hard to work with, and we have to remember that this is a long term game and the way I think about investing is would I not only want to work for you, would I not only want to take your call at midnight, but would I'd love to work with you again and again and again? And you should ask the same thing.

Michael Krigsman: Founder of the investors Ed, if entrepreneurs want to pitch you, what should they do?

Ed Sim: Buy me on X I guess. X.com/edsim, go to my newsletter www.whatshotvc.com. Find me on LinkedIn, you know, under Ed Sim or reach out to any of my partners or our boldstart VC X handle and we're open, willing. Just make sure you haven't raised any capital yet. We like these big ideas, and we like to be first.

Michael Krigsman: VCs are notorious for not wanting to meet with anybody that they haven't had a personal introduction from.

Ed Sim: So we do take some meetings, not a lot because a lot of times we get AI driven emails and spiel. That is, they may not even have the wrong firm name. They might have the wrong firm name in the email. So make it personalized. That's, once again, do the little extra work. This is a people game. No matter how much AI is going to be out there at the end of the day when you make an investment and when you have a partnership or when you sell a large enterprise deal, there's a relationship that gets built; there's a trust that gets built. So no matter what we do, let's not forget that human nature because it's never going to go away.

Michael Krigsman: Using ChatGPT to develop spray and pray emails to VCs is not the road to success.

Ed Sim: We've had so much more of that lately than we have before and so it's easier to stand out in a very personalized manner.

Michael Krigsman: I get pitched all the time from companies who want to be guests on CXOTalk, and it's exactly the same thing.

Ed Sim: It just goes back to the basics, right? I just say, hey, look, as much as this gen AI AI craze is happening, it's the same thing we saw back in 2017. The same thing I saw on the Internet. Boom, back in the day. It goes back to, what problem are you solving, how are you doing it ten times better? And who's going to pay for it, and how are you better than what's out there? That's it. We can complicate it all we want. It's just the technology. There'll be more technologies. And who knows if transformer technology is the right way to go. People think there's a data problem, that we're running out of data or training everything on Reddit and then applying it to an enterprise won't work. There's going to be lots of opportunities to build new things, and there'll be a new, new thing five years from now, I've no doubt.

Michael Krigsman: And with that, I want to say a huge thank you to Ed Sim. He is the founder and general partner of boldstart ventures. Ed, thank you so much for taking the time to share your expertise with us today. I really, really appreciate it.

Ed Sim: Ed, thanks for having me again. And hopefully we can do this seven and a half years from now again to see what happened, because it may be another AI craze.

Michael Krigsman: Again, everybody, thank you for watching, especially you folks who ask such excellent questions. Now, the next thing that's going to happen is by Monday, we will have an edited version of this conversation on cxotalk.com, where we'll tighten it up a little bit. So check it out, and then we'll get a transcript made, and we'll publish it.

Before you go, please subscribe to our newsletter and subscribe to our YouTube channel. We have amazing shows. Coming up next week is the CEO of Unisys. So, check it out, everybody.

Thank you so much, Ed, and have a great day. Take care, everybody.