VC Insights: Vital Lessons from AI Investing

Yousuf Khan: We're constantly calibrating our BS barometer. Sometimes it comes down to the ability for them to articulate an answer, the level of pause that you have when you ask a relatively simple question, which requires just a little bit of sincerity, etc. I understand that people are trying to navigate this journey, and fundraising is hard, and presenting to VCs is hard.

Michael Krigsman: Welcome to episode 841 of CXOTalk. I'm your host, Michael Krigsman. On CXOTalk, we explore the intersection of AI, leadership, and the digital economy. Today, we're discussing the impact of AI on venture capital, looking beyond the hype with important implications for both the enterprise and startups.

Our guest is Yousuf Khan. He's a partner at Ridge Ventures and a former multi-time chief information officer. What are the implications that AI has had on VC investing? What's going on with that?

Yousuf Khan: One way to break down the AI landscape is, as I'm quoting my friend regularly and basically Capital says, "There's NVIDIA and everybody else." That's brought to fruition a number of companies in the semiconductor space. And now, that's become much more hot than it was before.

It was very nice. And that's where Cerebras Labs went public recently. Then, one layer deep is really the distribution layer. And so you have the cloud providers like AWS and Azure and GCP. I'd probably bring Oracle into that. Why not? And then on top of that, you base it on the foundational models. And that ranges from what OpenAI or I've been doing to Anthropic and others.

And as a result of all of that together, you really have applications basically being built, both in enterprises and companies being built on top of it. Now, to your point, you said, "Well, there's a bunch of AI things there. I don't see them very effective." And there's a very good reason for that. They are what I term as wrapper companies.

It's a wrapper that is sitting on top of, say, OpenAI. They're probably building a very good product. I'm not sure that they will be great businesses. So it may be or may not be a great investable business. And the reason for that is AI models need to be trained. They need to be tuned. They need to be updated. They really need to be managed and secured.

And we haven't had enough cycles and we definitively do not have a talent pool out there. I mean, if you are building an AI company and I said, "Okay, well, I want to build an AI company for the longer term, how do you look to scale your model? Who is your head of AI?"

You would have a hard time. You'd be hard-pressed to find individuals. Data scientists have always been hard to acquire. Now, they're even more in demand. And so, it's actually very difficult to do that. So there will be companies, successful companies coming out of this. It really comes down to a core that people need to ask, "Are you solving a business problem?"

Are you solving a core problem? So, for example, the use of AI in medical research for, you know, disease prevention and research. Phenomenal use case, right? High levels of compute, data being infrastructure, that's a great use case. Demand forecasting comes up when I speak to CIOs and CFOs. You know, large this is these are Excel processes that have been going on for decades.

And so could that basically be shortened and executed? Most likely. Absolutely. And then there's going to be other cases where we're trying to place where it probably is nice. It probably is good, but it is not worthy of a venture investment because it's very difficult to scale. So that's the way I have kind of broken it down.

Michael Krigsman: So the wrapper companies, as you describe them, that are essentially putting on their own front end, for example, to the large language models, many of those will go by the wayside. That's essentially what you're saying.

Yousuf Khan: It will be first past the post AI system. I think the ones who have built a great product offering that will get traction, I think, in. But the fundamentals of building an enterprise software business still apply. How do you make your customers wildly successful? How are you able to keep your product at the cutting edge?

How are you able to make sure that it is secure? How do you make sure that your product being bought by businesses can scale? Right. So those apply with cloud and, and other, you know, SaaS for example. That was a very different challenge. The challenge was about enablement, and it was about being able to bring people on board to be able to use it.

It was about integrations with AI products which is vastly different. I mean, how do you know that your model is not up to date when you have. You mean you need to have a difference, a bench to be able to manage that? You're dependent on large language models. And so that takes time for you to be able to actually understand and be able to manage overall.

So some will be successful, and the ones that will be successful are the ones who have understood that, that they've hired the team to be able to bring into place, they've understood the customer problem, and the focus is kind of required. A lot of these companies, they just lose focus. So, well, we're really good at being able to, you know, send out, you know, whatever it may be in terms of a process.

But businesses are looking much more broadly. That's the "I don't want to point solution." And that existed in SaaS solutions. They exist in cloud software and now infrastructure. And now it exists in AI as well.

Michael Krigsman: Where is your investment focus? I think that will help give us some context.

Yousuf Khan: As a firm, Ridge Ventures has been around for 15 years. We raised our last fund last year, a $180 million fund. We invest seed in Series A, and we invest in enterprise software. And for us, the core fundamentals still apply. And that is that, yes, it probably is a very good team. They are probably building a very good product.

And the question really comes down to is, are they able to execute for the long term and build a company of long sustaining value? And so whether it is an AI company or a software company, the fundamentals still apply. For example, it probably is a very good product, but doesn't necessarily mean it is a good business.

And is it a good business? Is it in a market that is sizable where it can capture that? And then most importantly, is it an investable outcome? Is that really possible? So, you know, yes. Are we seeing more AI companies come into play? Are we seeing more startups that are using AI as part of their value proposition?

Those are that seems to be very frequent, where a company has maybe been around a year or two, has basically thought about like, hey, I can actually really transform my product offering by spending time on R&D to be able to bring AI and integrate it, and that does make a difference. That makes the product much more attractive. Yes.

And so and then the other core fundamentals come into play as well. Tell us a little bit about this founder. Let's talk about the team. How can they basically execute. Who are you competing with? So those apply for any type of software investment. So we you know, we continue to see an increase in the level of companies that are professing to be using AI in their product.

There are a bunch of companies who are working adjacent, so we're seeing companies that are looking to do things like secure AI, in enterprises looking to increase the deployment of AI, in enterprises. So, with every technological shift, Michael, what we've always seen pretty consistently is that there's an entire ecosystem that's been built around it, around talent.

For example, you know, I guarantee you somewhere down the line, people are changing their LinkedIn profiles to be prompt engineers. That may be a discipline for the longer term. They may not be, but the reality is that, you know, new talent, new skills start to get developed, new companies get developed, new problems occur. And for those problems, whether it's enterprise or early stage, companies need to solve for those.

And that's the excitement. So there is a huge opportunity as a result of these technological shifts.

Michael Krigsman: I just want to remind everybody we're speaking with Yousuf Khan. He is a partner with Ridge Ventures right now. There's a tweet chat taking place. Ask your questions on Twitter using the hashtag CXOTalk. If you're watching on LinkedIn, then just pop your question into the chat. Guys, folks! Take advantage of this opportunity because you can ask him pretty much whatever you want.

So ask. Ask, use this opportunity and ask your questions. So Yousuf, is there much difference now between VC investing in the middle of this AI boom versus traditional VC investment in enterprise software?

Yousuf Khan: I think the fundamentals have not changed. I think you still. You know, there's a lot of big firms and reputable firms which have not made sizable AI investments. I think the big difference is this is that I think a lot of the AI companies have had to raise a ton of capital.

And the reason for that is to be able to, of course, afford infrastructure and talent and then really be able to take it to market. And so we've seen a number of these companies raise very sizable rounds at very high valuations, which, by the way, you know, happened for a little while in, in a few years ago, you know, whether it was cloud or otherwise.

So the trend is relatively the same. The reason that I think it's different is just this. I think the scale of investing in these companies and the size of capital they're basically raising, and so that is, it will be interesting to see it's still too early to tell. But what is very clear is that businesses, broken up enterprises.

You know, let's just go through the notion, like 105 million people signed up to ChatGPT in less than ten days. Okay. I don't know the last time we had that. It was the iPhone moment, so to speak, the level of adoption that went in. But the difference is the mindshare that was captured at, you know, in not just in the business sector, but and at state level, you know, federal level, at city governments in schools, you know, this, that that is real.

And so as a result of that, I think, it is a slightly different, different paradigm shift, but it is a paradigm shift. Second, it really requires a degree of actually filtering out to figure out what is going to be successful. And I think that is a challenge. You know, like, you have to go through and understand, you know, the differences in the models.

And the other thing is the pace. AI is probably much more public. You know what I mean? So like you had the cloud providers and they were coming up with new releases and new product offerings all the time. It wasn't as I think we just live in a much more interconnected world, both with social media as well as a much growing audience and the best talk show CXOTalk.

And so therefore, as a result of that, the level of information sharing and then therefore the adoption and the learning continues to sort of happen. And as a result of that, you know, so when somebody releases a new model in OpenAI, it's all over. Right. Everyone's like, well, but trying to figure out whether this is the right thing or not.

And the pace of those changes also brings a little bit of a difference. So you may be looking at a company now and saying, oh, well, you know, tell us what you're basically building on. And then two months later, someone else brings in a model and says, well, by the way, this completely disseminates anything that you are thinking about.

So that gives us a little bit of room for caution. And so therefore, at least from an investment standpoint, you want to invest in the teams you understand that can relate to that.

Michael Krigsman: So you raised a very interesting point about the rapid growth. As you said, when ChatGPT was released, it had, you know, 100 million downloads or whatever. It was almost overnight. And it's changed popular culture as well. So given all of that, let's talk about hype. Where does hype fit into this and how do you ensure that you don't invest your investors' money in nice talking stuff that is just riding the wave but doesn't have legs?

Yousuf Khan: Michael, as you know, I wear a sweater vest to work. So the word "hype" doesn't... there is an industry term called hype cycle, which, is been very common with the analyst. There is a bunch of hype. I'm not going to deny that. There is a bunch of hype.

You know, I think there's two aspects of it. One is, if I were still a CIO, heaven forbid. And I was actually, bringing in technology. Yeah, I'd probably have to think the, the, the way my perspective on that would basically be. Well, I'm bringing in technology to solve a problem and I can do it a lot faster.

And so therefore I'm not going to look at it from any other perspective. I'm like, it makes, you know, the team look good. It solves the core business problem. And it's basically it's a modern approach on the investment side,

either fiduciary duty to our investors. And so we pride ourselves, as a VC firm, to do the due diligence that we need, before we make the investment, there will be hundreds of companies and there will be hundreds of founders, and everything will be okay in the world of venture capital in the longer term.

And so what actually is required is patience. We don't want to invest in companies where, we generally think, you know, will be problematic. I think nobody wants to do that. But in order for you to actually get to that decision, it requires you to ask your question. So for those who are would be founders who are looking back and, and thinking about, building company, the questions that, you know, most venture capitalists will ask are pretty much along the same lines, because at this stage of investing, my it's really about the team.

And so we're asking, can they build a product? Can they build a company? Can they hire an assemble a team, a world-class team that can that they can execute a great strategy on? How can they sell to customers? You know, what is the can do they understand go to market. And that applies, and that's something that we're going to have to basically do.

The AI piece too. It comes in two fold. One is, is AI accelerator is part of that product or it is the product. And that, you know, both of those questions, the way they're answered, you have to basically dig a little bit deeper. So that's provided a little bit of a perspective to be able to get into place.

But it is still very, very early days.

Michael Krigsman: So if I can summarize. Essentially, you were looking for truth-teller founders who can actually execute and deliver.

Yousuf Khan: We would definitely want them to be truth-tellers. But, probably. And that very core of it, you know, are they able to, assemble as a team and be able to execute on a strategy that, that, you know, they believe in? Really, early stage investing is really a large part of, in my viewpoint, is investing in the team.

It's the level of conviction that they have that they're able to drive things through and to be able to make things happen irrespective of what's happening in the market or otherwise.

Michael Krigsman: And thank you for putting a clearer point on my flippant question. And I'll just say my own bias is having been involved with enterprise software for so many years, for decades, I have become so skeptical of the hype of, you know, whether it's ERP, CRM, whatever, whatever acronyms and buzzwords you want and everybody, everybody meaning enterprise software companies jumping on the bandwagon and startup founders jumping on the bandwagon.

And it sounds good, but you sort of rip the rip open a little bit, the veneer and this stuff is not interesting. All looks the same. And basically marketing hype.

Yousuf Khan: One of the things that you and I both have in common is we're constantly calibrating our BS barometer. And so, you know, in my line of work, as in your line of work, sometimes it comes down to the ability for them to articulate an answer, the level of pause that you have when you ask a relatively simple question, which requires just a little bit of sincerity, etc. I understand that people are trying to navigate this journey, and fundraising is hard, and presenting to VCs is hard, but ultimately we're really trying to navigate ourselves, to be able to discover really the truth and trying to get to the bottom of trying to figure out of making a sensible investment decision. It's very, very easy to be able to tell, at least in my viewpoint, when a team is not showing the signs of being a cohesive, high-performing team.

I have that vantage point because I manage a number of teams that I learned through success and failure what that looks like. And so I apply that, that CIO experience to how I, when I meet with, when I meet the founders and companies and the other part of that is, of course, I provide a perspective of, as a former buyer of technology, whether some of my CIO friends would basically buy that.

Michael Krigsman: Please subscribe to our newsletter and subscribe to our YouTube channel and tell your friends.

We have some questions on LinkedIn and Twitter. This is from Shail Khiyara and Shail says, "What is your view on the small state-of-the-art models versus the world microdosing on generative AI?" He says small models can indeed create a big impact.

Yousuf Khan: I am a big fan of the small models. My core things that I want to relate to is ultimately will come down to the application. It will come down to the use case of where that basically works if it's narrow enough from my standpoint. And by the way, narrow doesn't necessarily mean it doesn't, is not a large B2B use case or a large industry. Not at all. It's much more about is it able to be, you know, really high throughput, and high output. So I'm totally fine with that. It really comes down to the application, the use case, and basically the place.

Michael Krigsman: Arsalan Khan, who's a regular listener, he always asks great questions. He asks the following. He says, "If we were to deploy customized AI using an organization's data, how do you address the locked data in ERP and CRM?" So are you familiar with anybody that's really solved that problem?

Yousuf Khan: A lot of companies are trying to solve this problem, as a forcing function. And let me tell you why CXOs are typically looking at this, and understanding. And it goes back to my original point. If you just think about the level of mindshare that happened with the explosion of ChatGPT and what OpenAI had been doing, that forced enterprises and companies to ask a number of questions.

And so that basically meant that when they were talking to an interesting company to bring into place, they ask a question about how they're able to secure the data, how they are using the data. Are they able to basically are they using it, you know, to basically train models that anybody else benefits? So these questions weren't as asked as frequently as they were before.

That's happening to the question whether somebody has solved it or not. It's still very early to tell. Some people are trying to solve it as part of, again, how I mean, how they're looking to deploy a product out because a company is saying, by the way, I want you to be able to it and to be able to sort of bring in, bring it into a central source, which I have access to.

Others, having basically solved it, the one thing I want to point out is each not to that question very specifically, there are a number of problems that have appeared as a result of the AI that are being adopted into enterprises. There are new cyber threats, there are new threat vectors that come into place. Those need to be secured.

The aspect of, you know, privacy that trust and safety. These are now in the mainstream conversation, which you didn't probably have before. If you think about when you mine assassin ERP solution, you weren't really having those conversations, you would think about, okay, is my data secure? And that was pretty much about it. This is not the case now.

Right. And so as a result, there's a whole bunch of even for the large language model providers, there's a whole bunch of problems that have come into place. And those problems need to be solved either by them or by companies who can serve them. Right. So we are seeing, for example, companies that are able to protect against the in kind of a DLP equivalent, data loss prevention equivalent for AI where an enterprise is using, you know, has got ChatGPT.

But being able to not just give the data away by being able to give customer data and ask ChatGPT or one of these other models about a particular use case, that's it. You know, there's a number of companies building that, so they haven't I haven't seen anything specifically, but I'm of course happy to basically talk to companies that do.

Michael Krigsman: Gus Bekdash has a couple of questions. He's also a regular listener. And these are quite thoughtful. So Gus Bekdash asks. He says that "there are enormous opportunities in AI, but much of it is hard to monetize. It seems that AI is skewed in favor of the elite in terms of the resources and talent because of the costs."

As you alluded to earlier, Yousuf, can an average organization monetize these using or developing AI solutions, or is this just skewing to the elite and the rest of us ordinary people so just you know, bow out.

Yousuf Khan: There's a ton of opportunity. I think that that requires somebody to build a smart business model. There are no shortage of companies that are wrapper companies or will sit on top of a large language model. Will have a use case that will charge for that use case or it may be copywriting. It may be, you know, think about marketing operations or being able to look at supply chain order of mine.

You don't need to build a very large language model to be able to do that. That's a very use case that's very specific. You can build a sizable business, that could that enterprises are willing to pay for. So I don't completely agree with that. What I the premise, is then, yes, a large part of the core infrastructure and building blocks for AI is largely held in, in, in a number of a handful of companies.

Sure. I will add that, but their objective is to really be able to distribute that as well for economic benefit and as part of the system.

Michael Krigsman: And Gus Bekdash comes back with a second question, and then we're going to move on to some of the others from other folks. And he says "AI solutions are low value in the beginning relative to their potential, but can be high value if the solution is scaled and survive that low value phase."

And so what does that mean from an investment standpoint where you need to pump in all of this money at the start and it's very uncertain?And I'm really interested in this from both a VC investment perspective and startups, but equally as much in terms of enterprise investments. We're not talking necessarily an R&D investment where you anticipate this, this enormously long time horizon with an uncertain result. Enterprises want an ROI.

Yousuf Khan: I think you've hit the nail on the head. Is. It's right. So, you know, the business will pay for something where they clearly, clearly see an ROI that comes into place. To Gus's point, yes. They start off with low value. If they are able to scale, if they are able to make customers wildly successful, that will be a sizable business where customers will be renewed because the impact of AI, the productivity uplift itself, is very visible.

And it's very exciting to be able to see, and you'd be hard pressed if you didn't want to basically adopt that on the enterprise side. There is a lot of experimentation happening. It's important to remind everyone is, you know, when enterprises are looking to do this, they're looking to solve a bunch of problems, but they're also looking to serve their customers.

And that requires a lot of experimentation. That goes into place very vastly different to how it was in the past. I mean, you know, this is kind of it's a very different vantage point. It's still far too early to tell. But this is what I hear from CXOs that top down a lot of initiatives are happening, and the expectations are being adequately managed to be able to see whether it's worth it or not.

Michael Krigsman: And I think it's worth mentioning at this point that you were a CIO at a number of different, high growth companies. So your vantage point is actually, as an enterprise practitioner turned VC.

Yousuf Khan: Very much so. Thanks for that. You know, I think part of that was I was an early customer of several companies. I was the first customer of an enterprise. The company called Moveworks. I was an early customer in support of Zoom that we're using now, you know, so, from my perspective, as I think a lot of other CIOs, when they see a product that is transformative, much better and impactful, whether it's in SaaS software or it's in enterprise AI, they're going to basically spend the time to be able to do that. So I think the fundamentals still apply.

Michael Krigsman: And this is from Yang Zhou. And Yang says that, they are doing research on gen AI, the use of gen AI by students. And the early findings indicate that the AI market is wide open. As a majority of the students, about 60%, are willing to explore and try new gen AI tools. So for small niche AI tools developers, the opportunities are definitely there.

Yousuf Khan: 100% hundred percent agree. And I think the tooling around it has become a lot easier. So, you know, the ability for ChatGPT, to help create your own GPT. Like. That's real. I tried it in under a few hours. I was able to create one for an entire business process I was doing. The quality in terms of long term is where it starts to get tricky.

So, yes, the ability to create and build something is very much there. And that will be a huge level of adoption for that. But to be able to build something sustainable, you know, with anything requires support and requires maintenance and requires figuring out what the edge cases are, requires dealing with all the bugs. And is it your fault or somebody else's?

Basically, and that's where it starts to start to be root I my so but definitively the adoption is very much there. The excitement around edge is very much there. And yeah, the kind of imagination and possibility is very much there, no question.

Michael Krigsman: So, there is opportunity for smaller developers to create meaningful products and get and even receive funding.

Yousuf Khan: When there is a platform shift, platform is created for you to do. So. Look at the examples iPhone and the App Store. Right. No shortage of people who decided to build apps on the App Store as well as Android, the on cloud. You could swipe a credit card and you're able to create a product on a cloud platform, be able to actually have a distributed and put it in place.

So those still apply. And it still applies for this one as well. Probably not as much, to be honest with you. But still definitely applies.

Michael Krigsman: I think it's also worth mentioning that we idolize. And the news and the media chases after these very large, I start ups, but you can build a really, really nice, worthwhile, successful business that's smaller, that doesn't need tremendous amounts of VC investment.

Yousuf Khan: Just to be crystal clear, if I was to approach my head. But I would have that conversation. I think I'd be. It could be a very nice business that could be built. What I'm saying is, yes, there are a number of businesses that could be built which don't require heavy level of venture investment.

And I think it is an enabler for you to be able to really, build a company around it, whether it's venture capital or not. But you have a platform to be able to do that.

Michael Krigsman: And I think just to the point of the of Ying Xiao, who asked that question. There are niches to be filled and I just I just think it's an important point to keep in mind. And for founders to keep in mind, to choose how they structure and fund their company and don't.

Yousuf Khan: Shortage of that like that. And so, I mean, you and I have we're very close and early adopters of Zoom and deployed that out. There was a whole set of companies that they just all they did was build virtual backgrounds and sold them million dollar revenue businesses, substantial outcomes, basically. So there's not every business that has to be a venture backer.

But if you have a wonderful idea that you think that people would buy too, no shortage of opportunities. And it and even more so now, I think for sure.

Michael Krigsman: We have an excellent question from Jitendriya Jamadar, who asks, "How do you choose an AI strategy to invest in the companies with high growth potential?"

Yousuf Khan: We're investing at the early stages. And so for us, it's trying to have an optimistic viewpoint to see that's high growth. Right. And so the way to look at that is really in two sectors. One is what is the size of market like is this going to be a high growth business.

Second is we don't have enough data points when you're investing at Seed in Series A, you don't have enough data points, enough cycles, all of these companies have barely got the bare bones of making a product, and a lot of them have just barely figured out, you know, the product market fit. When you look at high growth companies that are using AI, the question really comes down to is, you know, is it defensible?

And that's really where you start to dig deep to think about the team that is around this, that is basically being able to build a business and to be, you know, dependent. That's why wrapper companies have a bit of a problem. Right? They will sit on top of a large language model, and then it comes on to a better product, an execution with a little bit more depth.

As result, their execution may just overwhelm that. So, I don't think I've answered the question directly because I don't really invest in high growth companies. I invested the early day. I invest in companies that I hope will be high growth. But that's how I do it. I view them.

Michael Krigsman: I have to say, from personal experience with these wrapper companies, I like some of these tools. I pay for some of these tools. But when I use these tools, and especially if I get to know the founders, I'm always thinking, I feel bad for these guys because all it's going to take is ChatGPT to add a little bit more capability and you guys are toast.

Yousuf Khan: The level of innovation that's sort of being put in play that is being deployed by a number of these large language model companies. Number one is can be overwhelming. I mean, it's no question about that. Right. And the second piece is, is that it's transformative. It's not there's not just some incremental push that comes into place when you redeploy something.

I mean, it is really quite, quite something that's impactful business. And so the successful businesses are the ones who are on top of that, number one. But, you know, with respect, I'm, I feel they're serving a business that don't, don't care about the models. They just want to have the core use case being solved for. And there's going to be thousands or hundreds of thousands of buyers for those products.

And that's totally fine.

Michael Krigsman: We have another really thoughtful question from Arsalan Khan, who says, "If AI is disruptive, why are companies just making incremental changes? And do we need a different model for the acceptance and deployment of AI that goes beyond the AI, IT folks and the traditional software development life cycle?"

Yousuf Khan: Number one, it's still very early, so you don't have enough cycle time. You don't have enough cycles, you haven't had enough reference points. Right? And so to be able to say, well, that looks really good. That looks really bad, we should do this like, well, you don't have enough reference points. So as a result, it's still very early for enterprises to be able to move forward.

Second and this is a very critical point, it's a much more emotive technology that you're bringing to enterprise when you are bringing in a SaaS solution that was about, you know, agility and enablement. Same thing with cloud, right? Being able to bring up applications when you bring up AI, that is, yes, productivity uplift, but it's affecting some of the job.

If you are someone in a marketing operations function or an analyst function in finance or something, and you're saying, well, I'm going to bring it AI to be able to help solve, you know, be able to do my work a lot faster. That probably means you're not going to hire a bunch of folks, right? Well, that's also and it may also mean that basically, as it basically improves, you may be out of a job.

Right? So there is going to be that this is the human factor to it. Here's the other point. It's a lot more emotive because if it goes wrong it could be pretty substantial. Look what happened with Google, right? Like they had to shut that down because of the just the release of that model was not very well received.

And the results that were coming out of it. Well. And so the visibility of it, emergence in the out, if you said, well, I'm going to bring a solution in place and I'm going to deploy it out for my AB testing or, email campaigns, just imagine if that basically had a misspelling or something nefarious in content because you didn't check the copy or it was basically auto generating it.

Right. That affects your brand. That affects your customers. Right. Canadian Airlines suffered that, right. Hey, they used an AI chatbot in. The refund was totally valid, even though they didn't really realize that it was. They'd put that into the model. So the blast radius and the impact of something which you put in place, especially with AI, is, is substantial.

If you change the opportunity object in Salesforce and make a mistake, like everything's fine, we'll we'll figure that out. It doesn't the case if you actually deploy this out at scale. And that's that that's why there's hesitation and being able to take incremental steps.

Michael Krigsman: I love that phrase. The blast radius of AI is so broad. Potentially. And therefore, if you are, if you're at a startup or if you're a customer, you need to be paying attention to that. If you're a CIO, for example.

Yousuf Khan: That's why there is the so much conversation in city and state and schools. Like, okay, well let's take this way that you have, you know, results that you want to come in for central examinations in, in and get a university. And usually it takes super time because it's human. And now you're basically doing it through it through AI.

What if that was wrong? What have you basically sent out the wrong the wrong results? That's going to be a little bit a little bit problematic, what have you, you know, so I think that's where being able like excitement. Yes. But hesitation understandably because you have to basically figure out where the controls are. And not every company is very good at being able to present that.

Michael Krigsman: Gus Bekdash comes back again. He's he's on a roll today, and he's asking a strategy question, and a really hard one, actually. He's saying: "data from a single enterprise is often not enough to create a great AI solution. Therefore, aggregation across multiple companies, multiple data sources is required, but who wants to share their data. So how can a developer negotiate this dilemma?"

Yousuf Khan: I would basically look at something like transfer learning and number one, and I would look at something like cement, using lot of synthetic data to be able to do it. This is why the precision model for the recall for a number of these models at 80%, in single enterprises, for me at least, it's a good first step.

Right? So, the and to the question about whether some companies would share or not, some would, by the way. And the reason it depends on the use case. Right. So yes, people are going to be sharing, probably their financial models. Probably not. They're going to be doing intellectual property stuff, but some bare bones, whether it's data loggers or others, to be able to look at some optimization in sort of infrastructure may well be, the, the core thing, the core sentiment, it comes down to the actual application use case.

The other aspect is, you know, unsupervised learning is now definitively in the mainstream and being able to deploy it. So actually, I don't think there is a movement which is not so much about requiring large sets of data. You actually have the ability to be able to use small sets of data to be able to build out a model and be able to combine that with multiple techniques, to be able to build a workable product.

Michael Krigsman: Let me mention an example in healthcare, where somebody is actually grappling with this. So, John Halamka, who has been a guest on CXOTalk several times, is the president of the Mayo Platform Clinic. They have built a data platform where they are gathering data from a number of health systems across the country, and actually maybe, maybe internationally.

And there you have an example where the data is intensely private, right? This is this is health individual's health data. And so what they've done is they've set up a federated system where each organization that's providing this data, and I should mention the purpose of this is to aggregate data for, running AI on diseases and from basically solutions to medical medical problems.

And so you must have the data exactly as goes back to was asking about. But it it ensures that the data owner retains ownership and control at all times while enabling the broader consumer or the researcher to use that large, heterogeneous body of data from many different institutions.

Yousuf Khan: It's important to say there are data marketplaces, right? There are the ability for people to be able to literally sell their data, to be able to, to make revenue accordingly, like that is, that is real to Gus's original point about enterprises. I totally understand where there is definitive hesitation to be able to do that for intellectual property purposes.

Some of it's regulation I my belief is that there's going to be more, data regulation about sovereignty and data residency, in countries in states, and there's going to be more enforcement of, of those data regulations. So it's time will tell. But totally, totally valid point.

Michael Krigsman: And going back to LinkedIn, Ying Xiao comes back with another question, and he is also on a roll here. And he says what about what he calls the gen A.I. equalizer effect, meaning "large language models may narrow the gap between or the the grade point average between students. For example, a student of strong academics may lose to a student with less book knowledge, but very strong using gen AI." In other words, Gen AI may equalize these two students, the poor student and the stronger student. Any thoughts about that?

Yousuf Khan: Yes. Plenty of thoughts. So first. First and foremost is, depending on the subject. Actually, the quality is actually not that great for ones which require reasoning. So I've tested this out. And I used to be CEO in higher education institution, so, I've had some, some experience in the ranges of plagiarism type technology, right down to, you know, authenticity and others basically attribution like that.

That was core part of of some of my experience. I see that, not I'm actually not worried about that on the, on on things that whereas English literature and being able to do writing on the citizen sky. Yeah. There's no denying that. Right. That's, that's absolutely the case. Is that an opportunity to be able to put in some degree of controls around that?

Absolutely. It's actually pretty easy to tell. I mean, I could tell anyway, when I get an email which is generated by an AI solution, it's it's pretty clear you can you can be able to to see that. And that's definitely well, that's one of the deficiencies on the academic side. Controls are probably going to be put into place number one.

And the adoption is actually kind of mixed, if I'm perfectly honest. It's good to augment, and maybe start to be able to frame something, but to be able to make it happen, I think there will be some sort of technology that that allows us to be able to, to decipher that in some way, shape, form.

Michael Krigsman: We have a question from Lisbeth Shaw. Who says, "What role does AI play in how an angel investor chooses which AI startup to invest in?"

Yousuf Khan: Angel investors will invest typically in the people. So, I don't, you know, angel investors, with respect, comes on the back of really investing in people, because that's the stage that you invest in. You may have a thesis. Most VC firms have built out an investment thesis and a strategy. For example, we want to invest in, securing AI models and what that looks like.

And that's our investment thesis. So I think it's much more, the way I, you know, really, it's really two areas. One is an individual or team that they want to want to back because they think they could be building an amazing company. And the second is they want to invest in very specific areas. I would typically, you know, that would be the only two, things that I would recommend if you are doing angel investing is just know what you're investing in at the very basic level, it sounds so fundamental like, but you'd be surprised to be like, oh yeah, I'm just gonna invest.

That sounds like an interesting space, is it? Not sure it is. Tell me why it's interesting. And then go a couple of layers deep. You know, it's important.

Michael Krigsman: I have to imagine also that for angel investors as well as VCs coming back to the hype that we were talking about earlier. When you meet a founder who is so persuasive, there is a human tendency to want to believe and to therefore fall prey. I mean, many examples come to mind. I mean, look at Theranos, look at WeWork.

Yousuf Khan: Lucky, I work with a good team of people. Alex Rosen has been an experienced investor for a long while. Actually was a former founder herself. So, you know, I think it's important to be able to be in a collaborative, collaborative environment that basically suits me.

You know, actually, Tim Draper, the famous VC, I spoke to him some time ago. He said some of the best deals are the ones are the most controversial. And by that, he meets with the ones that probably had the most level of debate. But his core point was that you are basically debating it through no shortage of stories where you've had a compelling, storyteller and somebody who is a presenter.

Ultimately, at that stage, you go back to core principles, which is what is this person's track record? What are they awesome at? What are they not so good? And then the other thing is, are they building a company that will be phenomenal. So that's I, I don't get fazed by it. At least that's what I'm, that's is how I, I'm more of a skeptic from that at that point.

Michael Krigsman: I want to invite everybody right now to subscribe to our newsletter. Subscribe to the CXOTalk newsletter so that we can keep you up to date and subscribe to our YouTube channel. And please tell your friends. Tell your colleagues about CXOTalk, because, you know, it's pretty darn good. And we have really good guests. All right. Thank you very much.

We have another from Jitendriya Jamadar on LinkedIn. Who says, "Do you think ChatGPT, Copilot, and Gemini will soon replace the Google search engine?"

Yousuf Khan: I think Perplexity has a better chance of doing that. I think they have done a phenomenal job in what they're basically building. I don't think so. And I'll tell you why people have said this about no shortage of technologies. So I may be dating myself, but, you know, there are still a lot of legacy tech that is that you consider legacy technology that is still very much around and doesn't go away.

There are people who will be attached to a way of working changes very hard for a bunch of folks. Do I think they'll take more mindshare? Sure. Let's agree that they I think they'll do it at a pretty high, high pace. Yeah, but changing one's behavior in some areas. Remember the consumer experience is very unforgiving.

Like somewhere down the line there a lot of these new tools coming in place. And you like using it you know and then you know, but Clubhouse, oh, that was all the rage. Right. Like, oh I want to definitely get on to that all of a sudden, you know, speak to someone on audio. Must be.

Yeah. Okay. Well, don't really use it anymore. I'm still going to go back on CXOTalk. My friend Michael, please. My day of the week. So I don't I don't think it's going to happen from that standpoint. I think I just the history doesn't show that.

Michael Krigsman: You were the CIO at a number of organizations in the past, as you described. You ran a CIO group that met regularly. You're in touch with lots of CIOs today when it comes to AI and investments. What are CIOs trying to do and what are the challenges that they face?

Yousuf Khan: Number one, they are concerned about how to secure the model. And any deployment that they're able to do, they, they just very wary of that. Number two, they're starting to narrow down on very specific use cases. Number three, at the top of the list typically have been things in operations and the operational efficiency. Number four, there is enough adoption and experimentation.

Going on that they want to actually do that for reasons of up-leveling their team and for talent retention purposes. So they're there's encouragement for that to happen because they can it's a lot easier to do. You don't need to do a six-month POC of ChatGPT. You basically are able to get your teams to sign up and to be able to start using it, have experimental and being able to do that.

And lastly, I think, there is concern about regulation, as well as at the general pace, of the deployment that the point that you made earlier is 100% correct. Like there launched this in place looks fantastic. And then, you know, you see another announcement coming to place like, oh, wow, that was a waste of time.

So there is optimism, but there is caution amongst that. That's the consensus that I typically tend to see with a lot of my CIO friends.

Michael Krigsman: Are you seeing pushback inside the enterprise for investment? Because, as we discussed earlier, the nature of AI investment is such that it requires tremendous investment upfront with uncertain results. And a whole bunch of collateral set of issues such as worker training, reskilling and so forth. So, again, to use your term, that blast radius is enormous. And that implies investment combined with high risk.

Yousuf Khan: Clarity of use case. There is plenty of skepticism out there for companies to say we're not an AI company. There's plenty of skepticism out there for enterprises to say we're not an AI company. We will probably use AI will happily basically do that. But we still have a business to run, and we still have earnings to report.

And we still have to make sure that we deliver to our customers. So, there is plenty of hesitation. It comes down to clarity of use case. If you are able to basically decipher that and multiple layers down, whether you are a founder or a CIO, but you're able to look at a business problem and to be able to look at all the nuances that involved, able to navigate what the problem areas are and how you look to solve them incrementally over time.

That will that is definitively those are projects and got invested AI initiatives as a general rule, are seeing investment, but not some. Not not. That's more for experimentation versus actual execution. That still takes some time to do.

Michael Krigsman: And finally, advice for startup entrepreneurs who are listening and saying, you know, they have the next great idea. What advice do you have for them? And also, how do they get in touch with you if they want to pitch you?

Yousuf Khan: You want to get in touch with me, connect me on LinkedIn. Happy talk. I love talking to founders. No problem at all. At general advice I give to founders is you have to have a high level of conviction. People ask me all the time, like, what is this, the right time to start a company? I said, if you're asking yourself, is the right time to start a company, you probably shouldn't be starting a company.

The best companies are formed with founders who, in my viewpoint, just have high degree of conviction. Are willing to, you know, figuratively break down walls and be able to basically make that happen. Whether it be Eric Yuan, who's founder of Zoom, deeply competitive space, video conferencing, when he first started that, he was, you know, experience engineer WebEx and built an entire platform, a solution which is now a monumental, successful company, Pure Storage, where I worked on a CIO.

It's not the first company that did flash storage, but definitively one of the best ones, because phenomenal engineer in John Colegrove and John Hayes is founders and it's phenomenal customer experience. So my advice is if you have high conviction and you want to be able to build an at least an enterprise software business, I'd love to be able to talk to you.

It's not always going to be a yes, but I'm happy to give advice and guidance and feedback. I just love talking to founders.

Michael Krigsman: I think oftentimes a no, while painful, can be more helpful than a yes. Where the yes is half-hearted. When you get to know that's that's there's clarity there.

Yousuf Khan: There is value in feedback. And so my assurance to the founders is I will be clear as to why, and I will also be clear as to my feedback on the product, because that's one of the things that I, I spent 20 years buying enterprise software products. And so I have enough data to be able to give that feedback.

Michael Krigsman: Very quickly, advice to CIOs and boards of directors and others inside large organizations who are looking at this AI landscape know they need to get involved and it's like a mess and a minefield. And what should they do?

Yousuf Khan: Biggest advice I would give, other than, you know, speak to Michael, please, ma'am. My biggest skill that need to develop in the C-suite right now is one of being able to communicate, a story and be able to drive change.

You have to practice and figure out how you are able to do that. When you have a transformational shift, such as AI, you have to have people that are excited, you have to demonstrate your excitement, and you have to demonstrate a vision to do that. These projects will fail when people are not excited about them. They will fail.

When there is no clarity of vision, there is there is failure when there's no clarity of purpose. And that comes from CIOs, board members, others who are able to very clearly articulate, very clearly enthuse and motivate individuals and the enterprise and their employees and their partners as to why they are on that journey. If you don't do that, it's not going to work very well.

Michael Krigsman: And with that, I want to say a huge thank you to Yousuf Khan for taking your time and sharing your expertise with us. Yousuf, it's such a pleasure to see you. And thank you for being here.

Yousuf Khan: Thank you. What you've been building for the community has been a phenomenal and a really very useful, useful resource of knowledge. So thank you very, very much.

Michael Krigsman: Thank you to the audience. You guys are so smart. You guys are awesome. And you audience make CXOTalk. So before you go, please subscribe to our newsletter and subscribe to our YouTube channel and tell your friends, you know, we have great, incredible shows coming up next week. We're speaking with the CEO of Zoho in a couple of weeks.

We have two of the world's experts on information, disinformation, and the impact on social media from the Clemson University Media Forensics Hub. We have a whole lot of shows coming up. Check out our website and check out our newsletter and you'll see who they are. Thanks so much everybody. Have a great day. We'll see you soon.