theCUBE + NYSE Wired: AI Agent Conference | Sarbjeet Johal, Stackpane & Simon Chan, Firsthand VC

Clips
More from theCUBE + NYSE Wired: AI Agent Conference

Simon Chan

General Partner

Firsthand VC

play_circle_outline AI Agent Conference purpose: bringing together top AI technologists and industry leaders

play_circle_outline Evolution of AI technology: from rec to autonomous software, multi-agent orchestrations

play_circle_outline Changing technology landscape: harmonizing data, agent control framework, agent-to-agent communication

play_circle_outline Challenges in implementing agents as a workforce: accessing backend systems, application data, process knowledge

play_circle_outline Role of human in AI: AI changing human roles, relationship between human agents and AI agents

Info
Transcript

Sarbjeet Johal, Stackpane & Simon Chan, Firsthand VC

Simon Chan

General Partner Firsthand VC

In this Mixture of Experts segment from theCUBE + NYSE Wired, theCUBE’s Dave Vellante sits down at the New York Stock Exchange with Simon Chan, general partner at Firsthand VC, and Sarbjeet Johal of Stackpane, to explore how agentic AI is moving from tech circles into real enterprise adoption. Chan shares why he launched the AI Agent Conference to bring AI technologists together with industry decision-makers across finance, healthcare, commerce and logistics – framing today’s moment as a supply–demand convergence where rapidly advancing models and agent frame... Read more

explore Keep Exploring

What is the goal of bringing together AI technologists and industry leaders? add

What are the current trends and discussions surrounding the advancement of AI technologies and their application in enterprise environments? add

What are the concepts related to the semantic layer and agent control framework in the context of investment strategies in venture capital? add

What is the role of MCP (multiple context protocol) in the evolution of agent-based systems, and what challenges do organizations face in integrating agents with backend systems and access to business process knowledge? add

What is the role of humans in relation to AI in the context of technological advancements and workflow integration? add

bolt Powered by CUBE AI

Sarbjeet Johal, Stackpane & Simon Chan, Firsthand VC

search

>> Hi, everybody. We're back at the New York Stock Exchange here at the Buttonwood Podium overlooking the options exchange at the NYSE. My name is Dave Vellante and this is our Mixture of Experts series, a continuing series bringing in entrepreneurs, investors, startups, technologists, practitioners, customers. I'm here with Sarbjeet Johal. My friend, good to see you.

>> Good to see you. - Wow, we were just up in Boston.

>> Now we come down to New York to meet. I love it.

>> We were at IBM Think, some good action there. And Simon Chan is here, former entrepreneur, investor, general partner at Firsthand VC. Good to see you. Thanks so much for coming on. First time on theCUBE.

>> Thank you. Yes, first time here. Great to have two gentlemen here having a conversation.

>> That's awesome. In our NYSE Wired and theCUBE studio. Okay. So tell us about your background and I want to get into the AI agent conference, and then we're just going to get into the market.

>> Sure. So my name is Simon Chan. I have been an entrepreneur for most of my life, I would say 17, 18 years. I've built a few companies before, sold some companies, and now I'm taking the easier side of the world. I invest in amazing early stage startup company through my VC fund called Firsthand VC. When I have spare time though, I like to bring amazing people together and we have just done a wonderful, exciting conference a few days ago. It's called the AI Agent Conference.

>> So tell me about that. The obvious question is why does the world need another AI conference and why on agents, and why now?

>> We do have a lot of gatherings, events, conferences around AI, and I think nowadays around agentic AI, right? With all the respect, these amazing gatherings, they're great bringing the best AI leaders together, but I think something is missing in the formula. If you look at what's happening in the history of AI, a lot of these technology, they were built for the tech companies. But what we are seeing here is with agentic AI, a lot of these technology are actually available for the wider industries. And I think this is what we're missing, and I think this is what is unique about New York. We have the industry in finance, healthcare, commerce, logistics and all this. So my dream, my goal was can we bring two sides together, like the top AI technologists and these industry leaders coming together to have a conversation? So that's the AI agent-

>> So those companies applying AI,

>> not just the technologists, but those companies deploying AI. It's interesting, you used to work at SAP, right?

>> At Oracle. - At Oracle, okay. But you know SAP.

>> Yeah. - So back in the day, here's my point.

>> Back in the day, it really wasn't clear that SAP was going

>> to become the ERP leader and then they got escape velocity, and they won the game. But so it was hard to make that bet. You say investing is easy, that's why I'm impressed.

>> There's a special formula.

>> Technology's harder, but my point is, if you could have picked the companies that had deployed ERP most effectively, those, the New York Stock Exchange or whatever, probably did really, really well. But that's what you're trying to bring those two worlds together, a two-sided marketplace.

>> Yeah. Now, I think the very interesting thing is do we have more supply or do we have more demand? Supply means technological advancement, demand mean applications. Do we need that technology?

>> Ability to absorb it. - Right.

>> And what we see is,

>> there's always each side pushing each other together. So you can't have just one side of it. And what you see, I just wanted to say what we see these few years, but in fact it's just these few months, the large language models, agentic AI technologies getting ready. So the supply side is really, really high. But now we're pushing the demand side. Okay, with this technology, what can we do?

>> What can we do? Yeah. - What can we do in our industry?

>> And so, the opportunity is super exciting, especially for early stage companies. In the past, early stage companies have a really long enterprise sales cycle. I mean, that's what Oracle's, SAP sales also are great at getting the enterprise deals. But increasingly I see enterprise companies going in, learn about, "Hey, what is out there in the market? Can we do something?" So this is supply-demand catching up.

>> I think line-up business started buying stuff, like in cloud, started with line-up business. First it was like rogue IT. Line-up business came in and then IT said, "Oh, we need to do this," right? The same thing will happen in the AI space as well. The line-up business is buying. And then, we heard that at IBM ask me anything sessions as well. So we qualified this thinking, that now the discussion is going up to the CXO level and even the board level. So it starts with the practitioners. When the new technology comes in, practitioners hop onto it, then line-up business starts picking it up, then it goes to IT and CXOs.

>> So what were some

>> of the themes you heard at the AI agent conference? Where was the hype? Where was the reality?

>> The conference, we're lucky to have really amazing speakers, like high quality, top tier AI founders who are at the frontier of these technologies. We also have decision makers in large enterprise companies coming together. I think it's very clear that we're moving beyond even what the industry was talking about, like rec, for example, the technology where you can retrieve information, generate the response, make it less hallucinated, like less crazy, and then give you the results.

>> Less crazy, I like that. - Into what people are

>> thinking towards autonomous software, right? In my workforce or in my function serving our customers, are there things that we can dedicate to this new coworker called AI agents? And when there's limitation, now companies are actually looking at multi-agent orchestrations, just like what you're doing in a company right now. You don't just put one person to serve the customer, but it's a teamwork. How can we do that with the new technology? And of course that's the direction. But when we go deeper into the questions, especially in more regulated industries, like finance and healthcare, and it would be like, "Wait a minute, what is the margin of error-"

>> You can tolerate?

>> "... we can tolerate? " How do we regulate it psychologically? Are our customers ready for it? I think it goes from just the technology questions to a very human-centric trust question. And it's fascinating to hear all this conversation.

>> If I'm thinking about how AI is changing the stack, I mean, it's changing everything, from silicon all the way up into the application. And I'd love your thoughts on this, Simon. There's two emerging really high-value pieces of real estate in that stack, in addition to all the other stuff that's changing. But there's two new areas. One is the ability to harmonize all that data, whether it's different formats and different locations, and different access methods. So that I know that when I talk about revenue, it's not ARR or NRR, or calendar year or it's fiscal year. I know what actually we're talking about. It's not bookings, et cetera. Harmonizing that data. Some people call it the semantic layer, but beyond what we think of as the semantic layer, true harmonizing, maybe knowledge graphs come into play. And the other is what you had mentioned, sort of what we call the agent control framework. The ability to govern those agents, to trust those agents, so that when they act, the data is high quality. And I didn't have to have all these humans doing a data pipeline to get it to high quality. And it's governed, and I know the right people have the right access. It can interact with other agents. So, long-winded, but that's our premise. What's your investment premise at Firsthand VC? How are you thinking about picking your bets?

>> Things are changing very, very quickly. I think one traditional playbook in the VC world is thesis- driven investing. We have a board meeting maybe with some potential customers of these companies. At the beginning of the year we sit together, "Hey, this is the thesis we will invest in for this fund," and the fund cycle we're talking about, investment for three years, for periods of time. And then we write that out. And this is my investment thesis.

>> So as an example, if I could interrupt, during COVID, it was everything. It was about hybrid work and remote work. And boy, if you double down on that, you better pivot.

>> So I think the VC world, because of how fast technology is moving, we become more agile. It's kind of like iterative process learning. Now the technology is here, listen to the demand. What do we need? Do we need some more guardrail around this or are we looking for other technological breakthroughs to catch this wave? But I don't think, especially as an early-stage investor, I invest in inception-stage, very early-stage companies. I cannot just bet on, "Hey, this is the problem we need next three months. " I like to invest in founders with the domain expertise. If you're building the infrastructure companies, what have you seen before? Do you have the ability or willingness to talk to customers, to understand the needs? Investing in a learner is, I think it's my little secret, lazy formula. I invest in a big market, but at the same time, if this founder has unique insights, customer access to healthcare or to a certain domain, I think it will make their life and my life easier.

>> Yeah, actually I happened to have interaction

>> with Ron Conway, these guys seed- funded Twitter, Google.

>> Legend. - Big legend. He's a legend, right?

>> So he said, we had a quick discussion and I asked a question like, "What do you see? " Because we have invested in thousands of companies and there are big hits, of course, most fail, but few big hits. He said, "It's team and timing. It's team number one. Then timing." So-

>> I have three teams. TAM, team, timing. And so, I want a big TAM. I want a team that I can trust. And timing is interesting, right? So timing, what does that mean? Is it the right time? But I don't want a 20 year. I had Jim McNeil in recently. He invested I think 15, 20 years ago, I love you, Jim, but wow, are you patient, in nuclear fusion. And it's probably another 10 years before that thing takes off. But that is a too long of a time horizon for me. I could live with seven, eight, nine years. I'm getting older, so maybe I'm a little tighter, but seven to 10 years is a reasonable timeframe.

>> Seven, 10 years is the reasonable time.

>> And what is exciting about venture investing, especially at this stage is things really change so quickly. I have invested in some kind companies at the first few weeks of the formation of the company, and then they become a four, five-month old company, but they're already the industry leader at that new category, because of the demand. Just give you an example. We had the models, agents, models, LM models, but last few months or last couple of months, people are talking about how these things talk to each other, which becomes the MCP, the agent-to-agent technology. Would I be able to envision that, oh, we need a big market about agents talking to each other? Maybe, maybe not, but is there a company for you to invest in five months ago? Probably not. But now everyone's going into that. But of course that's another topic. Now it's hot. Should I invest in it or should I not invest in it?

>> Right, vector databases, RAG- based chatbots, I mean, it's like...

>> Actually, I usually say that technology is like medicines, right? Every pill you take, there's a side effect, and then we take another pill to kill that side effect and it creates another side effect. So in other words, there are gaps. We leave the gaps. When new things come in, there are gaps, right? Now there's governance gap right now. And then agent-to-agent communication gap. We need enterprise. We built enterprise services bus in the previous sort of era of technology after a while. So we will always start scrappy in software world. I mean, software is very unregulated industry. You can cook it up as you want and there's no regulation of course. And then we leave the gaps on security and compliance. So then people come in to fill that gap. And then, when we are filling that gap, another big thing comes and we start again, more gaps. I mean, I've seen people at Cisco just filling the gaps. They create a company, invest in it, and then they sell the company back to Cisco. So many times I've seen it. Same happens with other companies.

>> Look, the history of IBM was fixing its own problems.

>> You were at EMC, EMC was notorious for doing that.

>> Yeah, same thing. - VMware was another one.

>> Yeah, that's a really good point.

>> And that's why I always go back to the basic,

>> we're not just building the technology. You need supply, but you also need the demand. The best entrepreneur understand this demand side, and that's the role of Firsthand VC. That's the dream of the AI agents conference. Hence the name.

>> Actually, I like the name.

>> Thank you. Yes.

>> People with firsthand experience, with industry experience coming in. And with that, now you start to see the roadmap, a little bit of glimpse of the future roadmap of where things will happen. Now if we look back, we got LLMs or generative AI. Oh, you can generate very interesting creative writing images, videos, and now your enterprise customer starts to ask the questions. "Okay, this is super interesting, but wait, we want to use it in customer service. " You can't imagine a cue point code and send it to our customers. It will create problem. So we need some guardrail with our existing knowledge base. You generate things, but with that framework in mind-

>> With the error. - ...

>> so that now you create a rig things.

>> The next step, okay, you can do this, but can you automate some of our tasks? I think it will be really relevant. And the technology side will say, "We can't do it right now, because you are using so many other systems. You're using ServiceNow, you're using Salesforce, you're using this system. How can we talk to your system? " Now it comes to all this protocol, where agents talk to agents. You see, if you listen to your customers, the enterprise needs, you start to see where things will be you need.

>> So I have a question for both of you guys. So MCP, multiple context protocol, right? It stands for, an entropic standard, essentially so that agents can get access to tools and talk to each other. Now, my understanding is that we're trying to figure out how will this agentic wave evolve? Because you've got Salesforce doing its thing with Agentforce. You've got companies like Snowflake, for instance, bringing agents to where they are, Workday, ServiceNow, they're all doing it within their own SaaS environments. We heard from IBM, they want to be Switzerland of agents to do integration. We see Celonis suing SAP, because they can't get access to the data. My point is, to really have agents as a workforce, you need access to the backend systems, the application data, the logic, the business logic, and even the process knowledge, I would argue. And I don't think MCP is necessarily in and of itself going to solve that. Getting access, especially to that process knowledge, is going to be evidently challenging. How do you see that all playing out?

>> I think one element that we as the industry have not talked about enough is the involvement of human. There are certain areas where only human have access to. There are certain elements that you want humans to grant control, grant permission at the right time. If we take a step back to look at all this technology, technological breakthroughs, you see that we, you talk about the height of AI. What is real world is not real, right? Okay. We always think that AI will replace human. We have been talking about that for 50 years, I believe now. Same things for agents here, but if you look back to the history every time AI is not replacing humans. AI is changing the role of humans. Like, AI helped you do this, now your role has changed. Your role is to incorporate that into your workflow. I think it's the same thing here. The relationship between human agents, the knowledge, the workflow, and the AI agents. How do you put these two things together? So I would say if you need to build a good product for enterprise customers, think about, okay, we have this knowledgeable human agent with a more arguably trustworthy access to data. How do you put them into the picture to combine this to make it an integrated workable solution?

>> That's good advice. - Yeah, I see it a little

>> differently. So let's say there are five steps in a work stream, in a workflow. Somebody's going on a vacation, the system has to touch four or five systems, like HR, payroll-

>> Supply chain. - ...

>> other supply chain, all that stuff like that. So they can be their voicemail deactivation or whatever. So there are three scenarios, I think, in every process. One is it can be autonomously automated without any human supervision. The second is that human is in the loop. So two steps are done. Now an email goes to a human and they say, "Okay, I need to approve this or reject this, whatever. " And then, other three steps or four steps are done by the machines. So human is in the loop. And there's another thing that people are talking about, that I know, actually, I think it's a good way to describe it, human on the loop. So there's automation, but humans are watching it if it's doing the right thing or not. If it's not, they have ability to revert it. So I think those are the three options. But having said that, to your question, the agent-to- agent communication technically is feasible, easy to do thing. But in business terms, it's the most cumbersome thing to do. We have seen this from DLL health to API health, to now agent health. It will be the same thing. We're going to fight this-

>> Yeah, yeah, governance and things,

>> yeah. - ... and then the same thing. We used to have the enterprise services bus to settle all the stuff and roll back, and all that stuff. Ability to roll back. We will need that, but doing it across multiple vendors, like you're using SAP here and ServiceNow here, and Workday here, and then you're trying to make all these agents work with all these systems.

>> So that's music to your ears, because that is the white space that needs to be solved.

>> New companies will be formed in that area, yeah.

>> Last question. So I get a lot

>> of questions from young people. Maybe they're in college or they're in their first job. I want to be in DC and I always tell them, "Unless you're crunching numbers, you got to get some operational experience. Either get into sales or get into product. And if you're technical, build something. If you're quasi-technical, go into product management. If you like to sell, go into sales, get that operational experience. And then maybe you'll fail. It's okay, you'll learn from the failure. Maybe you'll get lucky. Ride a rocket ship, then you'll be golden. But get that operational experience." Is that viable advice? What else would you tell young people?

>> I think these are all great advice.

>> I've never thought of being an investor when I was younger. And as I get into this industry, let's call it a professional industry, being a VC, being an investor. One thing I learned is it's so commoditized. When I was building my company, we can easily find a unique value proposition, like I'm going after enterprise, mid-market, whatever. It's easy to find at least the initial unique value. Now VC, which VC doesn't tell you, "We're founders friendly. " No one's telling you, "We're founder hostile. " "We're founder friendly, we have good resources, we will support you"

>> Great network.

>> "... great network," everyone talk about the same thing.

>> You can really use ChatGPT to generate a pitch for you as an investor. Now, if you put your product hat, product designer hat on, the thing is, what is your uniqueness amongst all this commoditized VC resources? Why the founder will pick you? And so, I think the answer is actually more nuanced than going to do one thing, is to be unique, uniquely helpful to founders. So it could be like, you're a good networker, you have good insights of industry, you're a good number person. I don't know, but be unique. Offer some very unique values.

>> That's great. Awesome. Well, Simon, first of all, best of luck to you.

>> Thank you. - Really appreciate you coming on theCUBE.

>> And Sarbjeet, thank you.

>> Was my first time here. - Good to see you, finally.

>> Our friends in the CUBE Collective are going to have FOMO,

>> but we'll get you on, I promise. All right, thank you for watching. This is Dave Vellante and you're watching the Mixture of Experts series with theCUBE and NYSE Wired. Thanks for watching.