The Past, Present & Future of Generative AI—With Joanne Chen, General Partner at Foundation Capital

Richie Cotton: Welcome to Data Framed. This is Richie. AI companies are sprouting like weeds, and it's difficult to keep track of all the players. That causes a problem when trying to implement an AI strategy for your organization. To make any sensible decisions about a technology stack, you need to know what tools are available, and how they might fit together.

Fortunately, having an encyclopedic knowledge of what new startups are doing, is part of the job description for a venture capitalist. So, I've invited one on our show. Joanne Chen is a general partner at Foundation Capital. She's been involved in a wide variety of seed stage investments in AI startups, targeting infrastructure through to consumer facing products.

I'm keen to hear her views on the AI ecosystem, how she thinks these AI startups are going to affect everyone else, and how businesses can get the best value out of these new AI toys.

Hi, Joanne. Thank you for joining us on the show.

Joanne Chen: Thank you for having me. Super excited to be here.

Richie Cotton: Brilliant. So I guess to begin with, one thing I noticed is that there are so many new AI companies around, particularly generative AI companies. Do you have a sense of how you might go about categorizing them? Are there some important areas that are developing?

Joanne Chen: Sure. So from our standpoint, we've been investing. In applied AI for the last almost 15 years now. There's three areas that we typically look at broadly. One is the application of AI, so what we call intelligent automation, meaning automation of workflows, tasks, decision making. Think of applied to marketing or sales or to an industry like legal or logistics.

That's one big bucket, or FinTech, that's one big bucket. The second is machine learning infrastructure, which is all the tooling and all the, infrastructure necessary in order for companies to capture data, clean data, process data, and use that to leverage machine learning and AI.

And I think the third bucket, which I think is super interesting that we are spending a lot of time in is looking at cybersecurity, because as the tools get more powerful, as AI gets more powerful tools that are misused are, similarly, are getting more powerful. And therefore, how do you counteract things like fake content that's easier to generate and less costly to generate today?

How do you counteract hackers are now trying to do model poisoning or data poisoning? how are you defending against phishing attacks that are now sounding? more realistic, right? You get an email saying, please click this link and et cetera. So those are, there's another really big bucket that we're excited about.

Richie Cotton: Okay. So automation and infrastructure and also cybersecurity. These all seem like really important topics. the details of some of these later on. But are there any particularly important business use cases that you've seen have just emerged over the last year or so?

Joanne Chen: Business use cases. So I, maybe I'll comment on the application side. What I've seen companies pursue from a strategy standpoint there are two main things that stand out to me. One is around A time to market strategy, meaning when you are competing in, let's say, marketing you're building this new AI solution where you can offer something that hasn't existed before, if you're competing in like a really noisy sector, like marketing or sales, the companies that are fast to market or first to market and are, going about the land grab strategy, I think have an opportunity to Get those customers and build cheaper.

And what they will do is they'll build workflows, et cetera. That's like one broad area. And there's some other opportunities today that's touched very lightly by AI. So for example, procurement is one where there hasn't been that much AI innovation or innovation in general.

And so I think from a competitive standpoint, you can still go and build a better product and compete there, land grab, and then build deeper. That's like one broad I've seen in the app side. The second is like rethinking business models. So for example, a legal firm right now charges an ungodly amount per hour for, lawyer services.

And, their revenues are scaled by the number of lawyers they have. But with Edingener of AI, what you can imagine is a firm that just has 20 lawyers that are making a billion dollars in revenues because 80% of what they're doing is using machines and the 20% is all about customer experience and customer service.

And you can change the dollar per hour billing model to something entirely different. And I think so there's opportunities to reinvent certain industries by completely changing the business model thanks to this technology.

Richie Cotton: That's absolutely fascinating. I think anyone who's ever been charged hundreds of dollars just to be sent some kind of standard, like. legal form is going to be very happy if there's innovation the legal industry. Are there any other industries or sectors that you think have been particularly affected by AI or by generative AI in the last year?

Joanne Chen: But we're seeing innovation up and down the stack in all the three areas that I mentioned. There is also a big bucket of engineering automation or augmentation that I haven't mentioned. And I think this This especially applies to the engineers or maybe not the 10x engineers or even like the Silicon Valley engineers, but the engineers are doing more mundane stuff like integration building or kind of SRE.

So like thinking about site reliability, like the engineers are doing all the other stuff. And there's a ton of them. I think there's a lot of opportunities to augment and automate some of the work that they're doing and companies are more willing to buy software to augment or automate those pieces because they're less core to the company

Richie Cotton: Okay. So I do certainly like the idea of automating, like maybe the boring bits of engineering. That certainly seems a great thing to do. Like if you don't have to do all the boilerplate code, then that's definite win. Before we get into the details of like, particular roles, I'd like to talk a bit more about companies and how we go about finding good companies.

Cause I have to say just looking around at all the new companies cropping up, it seems like some of them are just like, well, What a website around the open AI API or something. It seems like it's not really like a a very substantial company, but there are some very cool things happening at the same time.

So do you have a sense of how you go about finding what are the quality companies out there?

Joanne Chen: for us, which, we invested the very, very early stage, meaning primarily the seed stage when companies don't have revenues, right? Like 85% of our investments are. Right. In companies that don't have revenues that haven't discovered product market fit. And so there's very little data. The way that we approach that is we take a very founder first perspective.

Like is this founder they have some kind of secret around the problem they're trying to solve. Do they have the persistence and tenacity to go about it? Especially because the ecosystem is moving so quickly and. What they built today could be irrelevant tomorrow. Do they have a initial team that is really high quality that can work together really well to move really quickly?

Those are some of the things that we really focus on. We do a lot of references. We, talk to people that they worked with before. We talk to potential customers and prospects. We have a customer network that we rely on. And so that's like one big bucket of investing strategy. I think the other big bucket is we spend time on problem statements that we think broadly have the opportunity to be reinvented.

For example, the, law firm example is one such. Opportunity, and we do this across the stack and look at where this new technology can make a really fast impact and then understand, like, what is the strategy they're pursuing

Richie Cotton: I guess at that early stage, you really are just investing more in potential than what's actually there already, it was pretty fascinating.

Joanne Chen: They don't, and then I think that's, it's a hard mandate to match the founders and the founding team to an interesting problem, but oftentimes we will spend years and years. Digging into a problem area before we invest, like, for example, Jasper, which is one of the companies we led the seed in we looked at marketing automation and marketing technology since 2014, try to invest in companies that were automating content creation in 2016, 2017 unsuccessfully, and then when we met Jasper Way before all this hype, it was just very obvious because we had been looking at that space for a long time.

Richie Cotton: And Jesper, of course, has been a huge success story for marketing content creation. Are there any other companies that you've seen that are particularly promising?

Joanne Chen: In our portfolio, we have companies including AnyScale, which is a unicorn company that is helping companies like OpenAI reduce their compute cost, right, with such large language models. There is a lot of costs associated with trading. This was the biggest innovation, in my opinion. Since 2017, when transformers were available to use as an architecture framework, it's really the cost.

And in 2017, we invested in companies that were trying to do some of this, but they did it themselves, right, versus use someone else's off the shelf models. And that was really, really difficult for a young company because you don't have that many resources. Right, so I think a lot of what OpenAI and others have done is like reduce the cost by doing their own training and offering that off the shelf.

Any scale is helping OpenAI reduce that cost. That's one example. Cerebros is another one in our portfolio. They're taking a hardware approach meaning they're offering GPUs. The way that NVIDIA does, and they're also offering models as a service for certain use cases. And in that case, the high level problem is still that compute is really expensive and how do we reduce that cost as quickly as possible.

Cerebrus's perspective is that a hardware approach is the right one, whereas AnyScale is, more of a software approach. Those are two other examples, but we have 50 plus companies in our portfolio that are leveraging AI significantly or are enabling the infrastructure.

Richie Cotton: And yeah, so both those companies you mentioned, so AnyScale and Cerberus, it's like really quite low level infrastructure stuff. So it does seem that... AI is causing changes at lots of different levels of the market, like from that low level infrastructure all the way up to applications and so on.

So, for organizations who are considering buying AI products or services, are there any things that you think they need to consider before they buy?

Joanne Chen: At the end of the day, whether someone has AI or not what really matters is like, are you moving the meter on a business metric? Business metrics are typically revenues and costs. And so when I think about the purpose of, AI technologies, it's really to serve a. and business problem or end user problem.

So I think that's like the most important thing to consider. I think the second is culturally, the balance between making sure someone's comfortable using AI as an augmentation tool, and how do you reduce the fear of full automation, I think that's something that's like, culturally talked about quite a bit.

My experience is such that no one has been fired because of an AI solution. But organizations have gone more efficient or can do more with fewer people, and I think that's a great thing.

Richie Cotton: Absolutely. And so that's originally mentioned the idea that a lot of organizations don't necessarily want a full AI solution, but they actually just want to make things more efficient, produce the number of humans involved. So can you maybe talk about the trade offs between having a full AI solution versus this hybrid approach with humans and AI together?

Joanne Chen: Oh, I don't think there's full AI solutions, period. I think most of the solutions are hybrid in nature because there's very few instances where you don't need a human interface layer. Whether that is a law firm that, AI can generate a contract for sure, to like some high accuracy today, but you still need a human to present that.

And provide the customer service. Similarly, if you have a sales team that's using AI to write emails and reach out, you still need a person to think through, am I prioritizing the right things? How do I make sure the customer is having a good human experience on top of, some of the automation?

So I don't, I don't actually see that many areas where there is full automation in the company. In fact, As the AI automation piece is commoditized, the role of the person becomes even more important.

Richie Cotton: This seems similar to your point about how you need to start with think about like revenue business metrics rather than starting with think about AI. So, do you want, just want to expand on what you were saying about how this customer experience is very important. So, do you start with trying to have goals around customer experience and then figure out how AI can be built into that?

Or do you have any thoughts on. How you approach this managing a good customer experience with AI.

Joanne Chen: I mean, I think that one of the things to explore, like for a startup, a startup has relatively limited funding. It can only hire so many people. And so how do you make sure everyone is ultra productive and, creating more opportunities and they usually can. So one, one of the things they could do is.

Build a stack of solutions where, if you're on the marketing team, use AI to create content, AI to do some of the lead generation and, and also help you understand where the channels that you should focus on from a, marketing spend standpoint, right? Like that, that can all be done by machines.

But then the person still has to come and think about the brand, think about like the interface between marketing and sales. Think about the community building and all the other aspects of it. So it actually frees up their time to do that. Plus all the, marketing specific things they had to do before.

Right? That's like when, what I mean by this kind of augmentation. But I do think one important question that founders are now asking is, Hey, like, let's assume we're going to adopt all these AI tools, right? To like help you be more productive. let's think through what is the importance of each person's role, right?

From a human standpoint, like, how do we start to quantify and qualify and how do we improve what we're uniquely best at?

Richie Cotton: So you mentioned that cybersecurity is one of the areas that you've been investing heavily in. And I think when it comes to security and data privacy, these are big concerns when it comes to AI. So, do you have any advice on how organizations can go about mitigating security or privacy risks when they're using AI?

Joanne Chen: so we have a number of companies that are helping with privacy as a big problem. I think it's going to be a problem for every single enterprise to solve. There's lots of different ways they can go about it. Everything from understanding who has access to what data within an organization to, maybe translating that data to something else that mimics the original data set to thinking about collection methods interfacing with their consumers and the whole set of issues around, around it.

But what that means is that organizations will need a tooling layer to manage privacy concerns and privacy issues. And I think that's a huge opportunity for startups.

Richie Cotton: Some of the companies you've been investing in infrastructure companies that are going to help it make it easier for you to build your own AI tool. But then there are also companies available that will sell you AI services. So like Jasper is going to sell you an AI service, but any scale is going to help you build it.

How would you go about deciding whether you build AI or whether you buy it?

Joanne Chen: Do you have good engineers? Do you have good engineers? Do you have the resources? Do you have, like, the organizational capabilities to prioritize this? It's the same way that, a lot of Silicon Valley large companies prefer to build in house because they have very strong engineers, and the rest of the world oftentimes prefers to buy because maybe they have fewer engineering resources, right?

Like, engineering resources is still a massive bottleneck, and the degree of that is different

Richie Cotton: Okay. So really it's got to decide, are you going to be playing to your strengths in one direction or the

Joanne Chen: if your strength as a company is something else, like you're selling insurance products, and like maybe your strength is your go to market team, then perhaps you will want to just, buy solutions for, for that to play to your strength.

Richie Cotton: In terms of if you are going to buy in some products or you're going to buy in some services then I think a lot of these AI startups, they're pretty new and maybe there's a high failure rate with new companies. So, do you have any recommendations on how you can judge the financial health or like the longevity of some of these startups that you might be wanting to work with?

Joanne Chen: Well, for us, we, in the business of funding these startups that have no money. So that is that is what we do. We have a. On the enterprise side, over a 75% graduation rate from, nothing to something C to Series A. So we help our companies increase the chance of success by helping them recruit, helping them get customers, shaping product market fit strategy, or discovering product market fit strategy, helping them with go to market and all that stuff.

So we try to increase the chance as a venture firm. As a large company, trying to... Buy solutions in this space. I mean, one of the things to look at is who are the investors behind these companies?

Richie Cotton: Okay. I suppose, if your company's invested, then it's a good sign. All right.

Joanne Chen: We're biased.

Richie Cotton: All right. So, I'd like to talk about like, the different kinds of generative AI that have cropped up. So I think chat GPT has obviously been like the, the most high profile hyped AI recently, but lots of different AI media types, things going on.

So, do you see GPT being dominant for a long time or are you seeing the adoption of other types of generative AI?

Joanne Chen: I think that chatGPT was a significant moment for the ecosystem. It was, we've been investing in all sorts of applications of different types of machine learning methods for a very long time. And the reason why chatGPT was still very, very significant is because. It was a moment where the mass market saw the powers of AI.

People who never could imagine manipulating in AI systems, who couldn't spell the word AI, can use it. So it democratized this technology, which was only accessed by very, very few people before, People who worked at Google, who graduated from Stanford, who had to study computer science.

That was the cohort who leveraged it, and now everyone can. That was like a significant moment, and that moment led to a lot of entrepreneurs, a lot of people, a lot of companies thinking about, now what do we do? Like, how are we going to make our businesses and our lives better by using this technology?

I think that was culturally a very significant moment. From a market share standpoint, I think that there is going to be larger fragmentation of available AI technologies to use, whether that's produced by open AI, which is still one of the leaders, but compared to two years ago, when we first invested in Jasper, for example, open AI has lost market share simply because a lot of other players have come in and offered a bunch of stuff Google or Amazon or startups.

And and I suspect that trend will continue because there's going to be, foundation models that are. More and more commoditized, there's going to be smaller models that are cheaper to use and optimized for different use cases. There's going to be proprietary models that enterprises will build because the compute costs will have gone down even more.

And that will be used for their businesses. So I think we're going to see a diverse set of models created by different people. And I think that's positive.

Richie Cotton: Absolutely. This huge increase in competition has got to be good for everyone, except maybe the people who were previously market leaders. Okay. So, one thing I noticed is that a lot of companies seem to be focusing on maybe text generation aspects or they focus on image or they focus on video.

And there seems to be this separation of different media that companies are interested in. Do you think that's going to continue or do you think there's going to be some crossover in the future?

Joanne Chen: Well, some of the techniques are different for the different types of data. But I do think when enterprises are adopting AI solutions, especially if they deal with different data types, so they will want a multi modal or multi... Type of, asset solution. What is still a bottleneck, I think, on the text, or not on the text, but on the image and video side, is that the compute cost for those is much more significant in text.

These are much richer and more expensive data types. And I think we're still at the forefront of some of the technology innovation for video especially. So we'll see likely the cost of, creating videos come down, especially on automatically. And I think the, chattativity point of that is still to come.

Richie Cotton: Okay. So maybe like the good video stuff is still in the future. So maybe we'll look forward to that

Joanne Chen: I mean, I haven't seen anything that can just generate a short clip on demand that looks great. Not yet.

Richie Cotton: Fingers crossed for the future then. All right. So, the other thing that's had quite a bit of hype in the last few months is the idea of AI agents. This is going a little bit beyond the traditional chatbot. So auto GPT, for example, got a huge amount of hype around April and May.

what do you think the future of AI agents is going to be?

Joanne Chen: It's a super interesting technology. I think there's a number of really early stage teams that are exploring use cases. Like I spoke with one yesterday and they were thinking about what is the vertical to apply this to. And I think that's still an exploration to be determined. My sense, though, is to offer an enterprise solution that accounts for a large part of automation.

They probably have to use agents plus machine learning, classical machine learning plus, workflow building and just integration building and all other types of technologies combined to really help, let's say, a real estate agent do his or her job holistically. Right. It's not a show us an agent can do something, maybe it can do some things better.

Prioritizing tasks, executing on tasks, thinking about the sequence of tasks, stuff like that. I think that's like really, really powerful with agents. But there's a whole host of other things, including what is the interface? What is like the integration to other systems that a startup needs to think through?

Richie Cotton: Okay. So yeah, I guess, just the ability to automatically send an email is a nice prototype. If you actually want to really automate someone's full job or maybe tasks, that's like, that's a pretty serious challenge. So, um, I guess we might be wasting

Joanne Chen: It's not that, like, I think the misconception is that this technology is going to magically give you a product in a very short period of time. I think it can give you a really interesting demo, but to have an enterprise ready product that accounts for a lot of other things that people care about in the enterprise, there's still a lot of building to do on top of what this technology is enabling, right?

The same, the same idea that, hey, if you're just a wrapper on top of open AI, you're probably not providing that much value if you're just a wrapper on top of something. Agent technologies, you're probably not providing enough value,

even though the agents are, you know, a little bit newer than what OpenAI has been building.

Richie Cotton: So can we talk a bit about the business models of some of these new AI companies? Can you maybe talk me through how these AI companies make their money.

Joanne Chen: I mean, I don't think they, they, they're making money any differently. The foundation layer, they're, charging on a per API basis for a lot of them. infrastructure. If it's a developer tool, maybe they're building some enterprise on top of it and charging as a subscription or some kind of usage dependent thing.

If you're like an application, most of them are charging on a, monthly or annual basis. So that hasn't really changed. We haven't seen companies that have evolved significantly where they can say, Hey, I'm going to change the economics of a law firm. instead of charging on a per hour basis, we're going to charge on a per product basis.

We haven't seen that mature yet to the point where that is actually happening.

Richie Cotton: Okay. So a lot of it is simple API subscription to things and maybe, do you think there's going to be any innovation in this space in terms of how these AI companies are going to change business practices or change, how they're monetizing things?

Joanne Chen: I think if the business model innovation is happening, then yes, I think the business can have a different pricing structure. But I, again, I don't think, AI businesses versus non AI businesses, that's not really a distinction in my mind, right? Like it's, the AI businesses are Using a cooler tool that maybe can reduce product development cycles, that can reduce go to market costs, that can make things more efficient, but it doesn't, it's not a business, right?

It's just one small, but very important part of, it's like, hey, you're going to war and you invented gunpowder. You have an advantage or the ones that are still, doing a hand to hand combat. But does that mean your war tactics don't matter anymore? Probably not, especially if the gunpowder is now open source and everyone can use it.

Richie Cotton: I like that open source gunpowder. Cool. So thanks. I'd like to ask you a little bit about economics as well. Cause my sort of view of this is like since maybe 2008 or so, we've had low interest rates and low inflation. And that meant that a lot of tech companies could delay monetization in favor of growth.

And that sort of changed last year. So now there's this big pressure to make money. I'm wondering how this is going to play out in the AI space.

Joanne Chen: Well, I think the high inflation and high interest rates affect the investment market first, first and foremost, and also everyone across the board. Right. So I think venture capital, for example, is scarcer today, and I'm not sure that's, my guess is not because there's not dry powder, but because there is just more.

hesitation to invest until people feel more comfortable that this company is going to be an enduring company. Whereas in 2021, everyone just invested very quickly because they knew the next round was going to get marked up because of a low interest rate environment. So that's trickling down across the board, I think venture funds are taking longer to raise venture fundraising cycles are also taking longer so all these transactions are slowing down, which puts the onus on quality, and I think that's a good thing because now companies need to think about, like, what is it that they're going to do to differentiate and make money and have a Much better product way beyond than just saying, Hey, I'm an AI company, right?

I don't think an AI company is anywhere sufficient, to get you funded.

Richie Cotton: Okay. So that's interesting that just saying, Hey, I'm AI isn't just going to get you that that instant sort of cash. All right. Interesting to know. So I'm also curious about the, partnerships between different companies. So, Microsoft have developed quite an interesting relationship with open AI is perhaps maybe the most high profile example.

Are there any other partnerships between companies that you've seen that are interesting? Thank you.

Joanne Chen: Well, I think the hyperscalers like Microsoft and Amazon and Google are all going to have their own research labs, if you will, like OpenAI is Microsoft's. Yeah. I'm being a little bit more edgy here. But Amazon will have the same, Google will have the same. Google has more of an investment there internally.

Amazon is partnering with a bunch of organizations. To do some of us in addition to investing in house so like the, folks that are selling cloud services that have a lot of budget, that have the ability to absorb compute costs, they're all going to be foundation model providers, And whether they're monetizing significantly on the, on these foundation models, or they are offering it because they want to sell other stuff in their suite of, things are, they're selling is TBD. But I, I would not be surprised to see more partnerships with any of these large players.

Richie Cotton: Okay. So, watch this space and see what gossip comes out of it, I suppose. So for people who are interested in getting jobs in the AI space do you have a sense of what sort of roles these AI companies are hiring for?

Joanne Chen: there's a couple of different categories. One is like folks that have actually trained large language models. There's very few of them, and that's why there's so much demand for like alumni coming out of open AI and others because they actually have that, capability. There's also just like what everyone was looking for before, which are like engineers or data engineers manipulating data, processing data, understanding how to.

Use data at scale. I think that's still a very high in demand job, and there's like other types of engineering work, which is more like, Oh, like, how do you put some of the outputs of AI in context? Right? This can be in the form of data analysts or even business folks that are using some of these AI systems.

So I don't think it has changed all that much from before, with the exception of a higher demand for folks that have trained large language models, which is a very small fan of.

Richie Cotton: Okay. So it's that kind of extreme machine learning engineer kind of position that's seems super in demand, but maybe there's only a small number of roles and beyond that it's the standard sort of data engineering, data science type stuff. So, related to this idea of roles, are there any skills that you've seen are particularly important?

Joanne Chen: Well, as, as more and more automation happens in a company, I, I go back to what does the person bring, right? So in my law firm example, if I were to build a law firm, I would hire 20 lawyers that are exceptional at customer service. And maybe that wasn't the criteria as much before because, they had to be skilled at understanding all these documents and stuff like that.

And they still do. Don't get me wrong. But like, What really, really matters from a competition standpoint is that the clients have to feel like they're amazing, right? So relationships, relationship management, relationships matter more than before. So I, I think that's a really important thing. I think engineers are always going to be a bucket of talent that's in high demand because everything in an organization is becoming somewhat technical, right?

Whether you're. Using an AI powered business solution, it's great to understand how it works and the limitations and how do you manipulate that and having that background, I think, is super, super useful across the board, whether whatever role you're taking

Richie Cotton: I really like that idea that customer support skills are incredibly important. It goes back to this idea you were talking about earlier. This is it's all about customer experience and that should be your, your big business goal. Okay. So, because Data Compass audience, we have a lot of data scientists, data analysts, and I've had a lot of questions recently about, well, saying, okay, I've got this role in data.

I want to get into AI. What do I do to transition from one area to another? Do you have any advice in this area?

Joanne Chen: and transitioning from one role to the other. So someone who's non technical, who maybe want to learn some of this stuff,

Richie Cotton: Oh, so actually these are a lot of people who do have this sort of technical data background, but then I was saying, okay, well. I've done data science. Maybe I want to try AI and what do I need to do to,

Joanne Chen: I think it's the same exact same skills. I don't think it really changes. I think what people need to do is be up to date on the latest. Innovations read research papers or have chat GPT help you read research papers and, and make sure they understand what's going on. I think that's like probably what's happening.

I think the folks that have studied data science and are practitioners are very well equipped and. Well suited to, be significant contributors in this new world.

I mean, it's just the same. It's the same transition. If you're a classical machine learning engineer and you're working with structured data to now, deep learning isn't is a capability like you probably have to, do some learning and reading around how to use deep learning and how do you manipulate, those systems.

But the fundamentals are still very similar, Like at the day, an AI is a prediction machine and you're taking data. You're creating something to predict something else. And so the questions like the methods are different in the case of tax, you're predicting what is the next sentence in the case of finance, you're predicting what could be the risk score.

and so at the highest level, like nothing has really changed.

Richie Cotton: that's good to know. I like that. It's just maybe a little bit of a different, different Python code, but a lot of the sort of similar ideas. before we wrap up, can you tell me what is your number one favorite AI company?

Joanne Chen: Ah, that is a hard question. I'm not sure I have a favorite. I am very excited for our portfolio companies. But I also think that I'm more excited, even more excited that just the number of people and entrepreneurs who are looking to start a company who are working on something interesting has really exploded.

And that's a net benefit to all of us and to this ecosystem. So that's probably what I'm most thrilled about.

Richie Cotton: All right. Brilliant. So the kind of the global benefit rather than any individual companies. I like that. It's a good good answer. It's going to avoid picking favorites. Okay. Is there anything you're working on right now or anyone you're investing in that you're particularly excited about though?

Joanne Chen: We are making a lot of investments. we have been investing in applied AI for the last 15 years and we continue to do that. So I think if folks are starting early stage companies, especially You know, at the formation stage we'd love to challenge.

Richie Cotton: Fantastic. And AI?

Joanne Chen: You've got to adopt automation and AI or you're going to be competing in a war with knives. that's going to be how it is, right? And right now there's still this huge bifurcation of companies that leverage AI and those that don't. Ten years from now, I think it'll be very rare to see anyone who still exists who's not leveraging that technology.

and so my advice would be understand your own capabilities, understand your talent pool, understand where, where the low hanging fruits that you can start working on and just know that it's either, you evolve and adopt and, figure out ways to compete or probably won't exist.

Thank

Richie Cotton: Nice. Okay. that's brilliant advice. Yes. Figure out where your strengths are and just figure out how to adopt things. All right. Excellent. Thank you very much for your time, Joanne. It was great having you on the show.

Joanne Chen: Thank you, Richie. Talk to you soon.