Building Category-Defining AI Startups with Yifan Zhang (AI2 Incubator)

We’re excited to feature Yifan Zhang, Managing Director at AI2 Incubator, one of the most successful AI incubators in the world, born out of the Allen Institute for AI and backed by a newly announced $80M Fund III.

Yifan is a 2x founder and CEO, Entrepreneur “50 Most Daring Entrepreneurs” honoree, and Forbes 30 Under 30 alum. Her journey spans building and exiting venture-backed startups, navigating extreme founder adversity, and now helping the next generation of AI founders go from idea to impact.

In this episode of From Startup to Exit, Yifan shares:

• How AI2 Incubator works differently from studios, accelerators, and traditional VCs
• What they look for in founders, and why ideas matter less than resilience and execution
• Why vertical, agentic AI applications are winning right now
• How AI startups are shifting from per-seat SaaS pricing to per-task, outcome-based models
• What it really takes to survive the founder “trough of adversity” and come out stronger
• Why Seattle may be the most underappreciated AI startup hub in the world

This conversation is a must-listen for AI founders, operators, investors, and anyone building in the agentic AI era.

Brought to you by TiE Seattle
Hosts: Shirish Nadkarni and Gowri Shankar
Producers: Minee Verma and Eesha Jain
YouTube Channel: https://www.youtube.com/@fromstartuptoexitpodcast

Transcript

Speaker 4 0:03

In this conversation, Yifan Zhang, managing director of A2 Incubator, discusses the launch of their third fund, their journey as a founder, and the incubator's unique approach to supporting startups. She emphasizes the importance of hands-on guidance for founders, the significance of market validation, and the evolving landscape of AI and entrepreneurship. The discussion also touches on the rise of agentic AI, emerging business models, and the global startup ecosystem.

Speaker 3 0:44

Welcome to the Startup to Exit podcast, where we'll bring you road-cost entrepreneurs and VCs to share their hard-earned success stories and secrets. This podcast has been brought to you by Thai Seattle. TIE Seattle offers a range of programs including the Go Vertical Startup Creation Weekend, Thai Entrepreneur Institute, and the Thai Seattle Angel Network. We encourage you to become a Thai member so you can gain access to these great programs. To become a member, please visit www.seattle.tai.org.

Speaker 2 1:30

Now, a word from our sponsor. At JP Morgan Private Bank, wealth is understood to be more than just numbers. It's about creating a legacy, achieving dreams, and securing a family future. With over 3.1 trillion in client assets under management globally, JP Morgan Private Bank is committed to providing customized financial advice that aligns with each client's unique goals. Clients benefit from a personalized approach, working closely with experts in philanthropy, family office management, fiduciary services, and special advisory services. The firm emphasizes building lasting relationships and ensuring that every financial strategy balls with the needs of clients and their families. JP Morgan Private Bank not only manages wealth, but also helps clients create a meaningful impact. Whether individuals seek to support charitable causes, manage a family legacy, explore new investment opportunities, the private bank team is there to provide guidance every step of the way. With the presence of key financial markets worldwide, JP Morgan Private Bank offers unparalleled access to global insights and opportunities. Its commitment to excellence and innovation ensures that clients receive the highest level of service and expertise. Time is recognized as precious, and JP Morgan Private Bank values every moment spent with its client. The mission is to make the financial journey as seamless and rewarding as possible, from the initial consultation to the ongoing management of athletes. To learn how JPMorgan Private Bank can help clients achieve their goals, visit privatebank.jpmorgan Connect with a dedicated team of specialists. Again, privatebank.jpmorgan.com. Thank you.

Speaker 1 3:40

Hello everyone and welcome. Today our guest is Yifan Zhang. She is the managing director for AI2 Incubator. It's a very well-known incubator in Seattle, has spun out a number of companies over the last few years. And they recently raised their third fund for $80 million. So congratulations. Thank you. So tell us a little bit more about the fund. It's quite a bit of a step up over your previous funds. What are your plans with the fund?

Speaker 4:12

Yes. Yeah, this week just announced fund three, our $80 million fund. It's actually almost 3x the size of our previous fund. Core elements of the incubator, I think, are not going to change. We invest $600,000 per startup that we incubate. This really allows us to do it over a longer time horizon. Our last fund actually we deployed over the course of about two and a half years. And this fund is going to let us deploy incubate over four and a half to five years and step up slightly the number of companies that we incubate per year. We're going to be doing about 15 companies per year versus 10. But all other elements of the incubator are pretty similar. We're just excited to be able to uh work on more companies. Okay. That's great.

Speaker 1 4:58

So let's talk a little bit about your experience coming in before joining AI2. You were a founder of a company called Loftium. A very interesting concept where you would allow home buyers to um buy homes and you'd help them with the initial um funding by renting one of the rooms in their home. Am I describing that correctly?

Speaker 5:22

Yeah, yeah, that's right. So we're an affordable housing, we call it a host to own startup, um loftium. We create a path to homeownership by um turning one of the rooms in a home into a short-term rental style, like Airbnb style income stream. Um so our renters can save up a down payment, get discounted rent while they're living in the home, and because of all these benefits, be able to own the home outright in three years. And so it's a you know a solution to some of the affordability challenges we see here in Seattle. And we were launched in 10 markets.

Speaker 1 5:56

Yeah, very interesting idea. Uh, how did the uh company do?

Speaker 5:59

Yeah, um, so we ended up being acquired by Fly Homes, another real estate company here in Seattle, um, at the beginning of 2023. This is actually Loftium was my second startup. Um, I've always been a founder before joining AI2 Incubator. I actually have never uh had a job. I went to Harbor undergrad and then started these two companies. You know, the it's been a very interesting ride. Loftium um was definitely, you know, we went through so many ups and downs, had raised over the course of our lifetime about 130 million uh between venture capital and uh asset back debt. And we actually own homes to be able to enable this host-to-own process to work, um, you know, expanded rapidly throughout town markets, had over 150,000 um, you know, renters in our homes. Um, we also uh had to survive the pandemic. We got hit really hard by the pandemic, which stopped travel. Um I have some crazy war stories I share with the AI2 founders um around going from uh essentially nothing like one to 30 million of revenue in one year, and then losing 90% of that in one week during the pandemic while I was fundraising and you know having to retrench, survive, raise another round and go through an acquisition. So being a founder is is wild. Yeah, yeah, that's great.

Speaker 1 7:18

At least you had a exit with a good exit with uh Fly Homes. So that's that's good. I really they're continuing to offer the uh Yeah.

Speaker 7:25

So at this point, Fly Homes has gone through a lot of transitions themselves in the last two years. Um I think you probably have seen some of the news. You know, they've actually sold off parts of the business and you know, retrenched in other parts of the business, mainly offering their cash offer through wholesale versus uh through their own brokerage. So at this point, I think Loftum is no longer the the core product, uh, nor are some of the other products that are before the last two years, which has been an interesting transition time for all of Prop Tech. Got it. Okay.

Speaker 1 7:58

So then you joined uh AI2 as a managing director. Yes. It looks like you've been an entrepreneur at heart. Uh how did you go about deciding to join AI2 versus starting a third company?

Speaker 8:11

Yeah, um, I mean, you know, like I mentioned, I had never had a job before. Um, I always thought of myself as a founder at heart, operator at heart. My first company actually was one of the top mobile fitness apps uh in the first wave of iPhone apps. Um so, you know, I had some options. The whole team had transitioned to fly homes at that point. I had some options. I was thinking about, you know, taking on maybe like senior leadership roles uh, you know, in in these larger companies and seeing the next stage, right, of startup growth. Or, you know, this really interesting managing director role at AI2 incubator came up. Um, I never thought of myself as a VC, but the incubator is different because yes, like half the job I think is as a VC, right? To evaluate startups, write investment memos, like have theses about where things are going in AI. The other half is very much hands-on with the founders. Directly with the founders, being able to give guidance and help shape like early product, go to market strategy, like customer profiles, all the way through to fundraising. And I think that was the part of the job that got me very excited. The, the hands-on, right with the founders. Also, I think being able to get an early stage right in AI was very exciting. Um, because I'd been through the early mobile revolution and and you know, had seen right what it's like to build in that early era and also the power of this platform shift, right? Um, I pretty quickly saw what was happening in the AI market and thought it was a very exciting moment to join. Yeah, got it. Excellent.

Speaker 1 9:49

So tell us a little bit about AI too and uh your model of working with uh founders. Um Pioneer Square Labs and Madrona Venture Labs when they existed, you know, had their uh spun out their own ideas. Uh, you know, they had a sp specific concept idea and then go find a CEO and then spin it out. Uh they also had a kind of model of just, you know, working with founders who had ideas and then they would help validate the idea and get it out to market and so forth. Um and they took a pretty hefty stake in the um in the company. What is what is your model?

Speaker 10:25

Yeah, um, I think this is the confusion between a studio, right, and an incubator or accelerator, which I use interchangeably. But I think studios, right, like PSL tend to come up with the ideas more, right? And because of that, uh they take a bigger equity time, right? Because they they have the idea and they're kind of building the team around that. We are much more an incubator or accelerator, so more similar in terms to something like YC or Tech Stars than than to the PSL scene. We have now decided uh we're we're not coming up with our own ideas. We need the founder to come with that spark. But we can start as early as a solo founder with just a seed of an idea, right? It doesn't have to be fully formed. We can help to to shape it. But a really important thing to me as a founder is that initial idea, right, typically will not be the final idea. Usually you go through so many iterations to get product market fit. And then once you have product market fit, the hard work is really scaling up that company. So to me, I I don't feel like uh, you know, the idea is the most important part. And also I think it's so important for the founder to have passion for their idea to be able to take them through all of the painful iterations, right? Ups and downs. I shared a little bit just from my loftium example, right? But every founder will go through this type of turmoil. And if you don't love what you're working on, there's no way it's not rational for you to stick with it to get to the success on the other side. So, you know, that that's a little bit of our thinking. Um, I'm happy to share like specific terms about the program too.

Speaker 1 12:03

Yeah, that would be great. Yeah, what what are what other uh the terms that you uh uh, you know, work with the founders.

Speaker 12:10

Perfect. Um, so we uh invest $600,000 and our terms are we invest that um as a safe note at a 10 million valuation. And this $600,000 can be split up between if you come in with a fully formed team, an idea, and maybe some validation, we can just invest that $600,000 up front. If you are earlier stage than that, which many of our founders are, right? You're just a single person, maybe you're a talented right researcher, maybe you're in a bigger tech bigger company um and you know uh deploying AI at scale, but you haven't um built up the team, right? And you just have that seed. We can start with our entrepreneur in residence program, which is a hundred thousand dollars, right, to every teammate, every co-founder, um, just up front. So you can start working, um, and then we'll put in the rest of that six hundred thousand dollars once you do have a little bit more of that teen technology or traction. On top of that, we're gonna terms for the 100,000. Sorry.

Speaker 1 13:07

Sorry, go ahead. What are the terms for the 100,000?

Speaker 13:11

Yeah, so uh same thing, right? $600,000 at $10 million valuation. You can break it up into $100,000 first and then $500,000. Everything's at the $10 million valuation, right? So eventually it's the same. Um, it just depends on what stage you're at when you come to the incubator. Um, on top of that, we do take a 7% common equity state uh for the incubation. And that uh is because we put in a lot of hands-on work, right, into the companies. Um, it's a six to 12 month incubation period versus, you know, a three-month accelerator like Techstars or YC. But the comparison, right, with something like YC is, you know, at the end of a seed round, the I2 incubator will own about 10%. Um, YC will own about 7%. So we are a bit more expensive, but I do think we're quite differentiated because you're not going to be part of a batch of 200 startups. We only incubate 15 companies per year between all of the partners. So you're gonna get a lot more hands-on attention and support. Got it, got it.

Speaker 1 14:14

Now, uh one of the things that uh, you know, MBL and Finance Square Labs uh used to do um or are do um is market validation. So they actually, you know, they'll create a prototype, you know, put ads uh uh out there, test responses, all of that to see if there's some and then do customer interviews, et cetera, along with the founder. Is that something that you guys also do?

Speaker 14:40

Yes, absolutely. We take a slightly different approach, I think. And and I I do think like some of the like prototype ads model work pretty well for the pre-AI era. Um, I think in the AI era, lots of things are changing, are being discovered. And I I think it does take a little bit more of a bespoke approach, um, depending on what it is that you're building. Like, for example, if you're building, you know, a like B2B, right, like vertical application for a very specific industry, like mortgage, right? Like it will not help for you to build a prototype and run ads at it. Like you need to have um, you know, feedback from very specific people within a mortgage originator, and you have to decide, you know, is it this type of originator that's bank or this type that's non-bank? So there's a lot of decisions that need to be figured out in the early stages. We very much are hands-on, right? Sitting next to our founders, helping them uh figure this out, calling out when customers maybe are saying one thing, but actually they mean something else. And you know, that that's more of our approach. Got it, got it.

Speaker 2 15:46

Okay. Uh over to you, uh, Gaudi. Great. Thanks. Um we start with this premise, right? You as an incubator have been on this journey much longer than the AI wave has hit the world in general. Seems like you're finally validated that you're on to it something much before the world woke up one morning to Chat GPT. But setting that aside, so you have significant experience now that you've done this many companies through your first two funds, etc. What are the general uh statistics of how many got spun out, how many got exited, how many funded, so that we have some sense of scale that you have achieved, because otherwise you there's no way there's only one comparison, YC, but that's not a that's not a real comparison because they they have a very different model as you as you Yeah.

Speaker 16:34

So uh, you know, uh some of this happened before I came and joined the Incubator, but I want to tell the team that we have here at AI2 Incubator, right? We swung out of the Allen Institute for AI, we've been investing AI for the past decade. I get to work with amazing world-class research minds, right? Like Orin Azioni, who founded AI2, Vu Ha, who's a founding researcher, like there are technical directors here. So it's I can't state how just helpful, right? Incredibly helpful it is to have them on my team when we're evaluating and working with AI startups. But in terms of metrics, you know, I think we're very happy and proud of our track record. And I think our new fund and the fast raise uh shows that. We have spun out at this point about 40 companies out of AI2 incubator. I believe the metrics are over 90% of our spin outs raise funding, you know, well ahead of YC, I believe. Um and I think the craziest stat is I think about a third of our companies have already exited, like actual acquisitions. So, you know, that that's like a pretty astounding metric to me. Um, and we've had some really major exits, uh, especially last year, that were public, right? We had Lexion here in Seattle, exit to DocuSign. We had uh Materia, which is an accounting and audit AI startup exit to Thomson Reuters. That one was a little bit more low-key. Everyone knows about Case Tex, the legal company that was acquired by Thomson Reuters. But Materia was a similar huge win. Um, and that was their bet in the financial side. You know, but we we have a number of these successes, and including many others that are more low-key and haven't done the full PR push in classic Seattle style. But we're we're very proud of the results for our companies.

Speaker 2 18:23

Wow. That's a that's a a lot, 90%. That that's a big uh, you know, big percentage to uh get funded. So if I'm an entrepreneur listening to you, could you break down um what you would look for in the founders, the idea, formate formation of that idea with the founder or founders? How would I say, hey, I think I'm a good candidate to go to AI2?

Speaker 18:50

Yeah, we've really clarified our vision for what we're looking for. And it's a little bit nuanced and differentiated. Um, so we're looking for two types of founders. One is a deeply technical, you know, AI expert. Um, so that could be someone who's a researcher, say at AI2 or, you know, other research institutes. It could be a PhD, like recent grad, even that has a very specific piece. It could be a technical practitioner in a big company like Microsoft. We work with like Microsoft Research, right? AWS, Google. Um, so we're we're looking for that technical expert. And we can start with that solo technical expert and kind of build right the team around them, or we start with a domain expert. And the domain expert typically comes from industry, right? So you could be an expert in I I mentioned mortgage, right, or professional services or manufacturing or fashion, right? But you have a deep um expertise in understanding how an industry where the gaps are that can be solved with AI, right? And most importantly, you have the network to be able to, once this thing is built, deploy and distribute very rapidly within usually more trust-based networks. So these are the two profiles of founders that will build companies around. And actually, the majority of our companies have spelled out, they met their co-founders while in the incubator. So we do like both of these profiles. Often people do kind of join teams with within the incubator.

Speaker 2 20:17

I see. So um uh given given given that profile, right, at this point, looks like you're building on some models that exist there. I mean, everybody is. So there's the if you look at the AI stack, there's the models at the bottom, which uh few players have, and then there's applications on top, or some are horizontal, and you mentioned some verticals. Is there a SKU bias at the moment to be vertical? Or are you still saying there's room for horizontal startups? I mean, where where is your thinking as an incubator?

Speaker 20:52

Yeah, so we are quite unique, and people are always surprised when I say we're vertical application layer focused. I think coming out of Allen Institute, right? People expect us to be like model layer or at the very least, like infrastructure horizontal, um, like tooling. But, you know, two years ago, actually, we had a conversation as a leadership to really think about, you know, the the future paths of AI, right? And where particularly early stage founders can play a crucial role. Um, and we've done all the layers in the past decade when things were not quite so competitive. And we still have like some actual like foundation model companies and horizontal companies. The issue is that our thesis is the the big model companies, right? Like two years ago were maybe like 70%, 65, 70% good enough at doing most tasks, right? Um, but there's certain industries like verticals where that is not good enough. Like you have to be at 90, 95%, maybe 98%, right? Um, like whatever you want to say, like accuracy, right, like quality, um, in order for the customer to actually purchase your product and continue purchasing. That product, renewing the product. So our thesis was that uh the role of the the founder, right, is to fill that gap as as high as possible, right? As close as possible to that like end like 95% goal. Um and the model companies would continue to like improve the quality over time. So it would go from like 70% up to 80%, maybe. Um, but it would be very difficult for them in these certain industries that have high standards to get up to like 98%, right? So there would be a gap for at least a while for these startups to create value. Horizontal plays um necessarily need to spread that value among many different industries and customers. So you could maybe get from like 70% to 80%, right? But it's very hard to get to 90% for every single customer because you are horizontal. Um so that was our thinking right around vertical application layer. At the time, people thought we're kind of crazy. You know, it was like a picks and shovels like time where a lot of horizontal plays were getting attention. But I think, you know, this year we've definitely seen our thesis prove out, uh, uh kind of, you know, graduated a number of very successful vertical application plays, right at Harper and Mortgage, KCM, and uh enterprise immigration in the professional services space. Um so we're we're feeling really happy about that choice and we're continuing to invest into vertical applications.

Speaker 2 23:22

Got it. So this year is the year of agentic AI, or at least it's being touted as one, right? Every every there's no company on the planet that isn't thinking agentic AI. Now, with your vertical focus and this spotlight on agencai AI at the moment, or at least the du jour of the day, is that marriage attracting different kinds of founders versus when you were saying, hey, we are vertically focused? Because it looks like they need this combination of domain expertise and technical expertise to come together to solve a particular problem. And the agentic AI will lend itself. So you think A, is this the year of agentic AI? B, is this changing the profile of founders that are coming to uh coming to you?

Speaker 24:10

So yes, I I do think it's the year of agentic AI, but I also think 2020, you know, since seven and eight will also be the years of agentic AI, right? I think I think there's a long way to go here. Um, and absolutely at the beginning of the year, like agentic AI wasn't working that well. It's still probably not working that well in most areas. So there's just a lot of work to be done. In terms of the profiles of the founders, actually, I think we set ourselves up perfectly for agentic AI because the way to actually make agents work is one you need actually deeply technical founders that can build this stuff. And it's hard. Everyone's figuring it out as we go, right? So um I think that's absolutely necessary to even build anything functional right now. Um, and then also the agents work better if they actually have context, right? If they're doing the right tasks. And also today, especially in 2025, uh, they're they're given very specific narrow tasks to do. Like in the future, I do think the scope of agents will increase, hopefully. But today you have to be very targeted because the agents are pretty preliminary. So having a domain expert that can know what is that very narrow but super high value task to be completed, um, that's hugely helpful, right? And what context does that agent need in order to do that job? Um, so yeah, I I think our thesis continues in the era of agentic AI, and we're you know doubling down on it.

Speaker 2 25:34

Got it. So, what business models are emerging as these two come together, right? Vertical plus agentic AI. There's a the SaaS model is under attack or at least under threat at this point, if not under attack. So is there new models that you're seeing emerge that you see founders gravitate or buyers gravitate towards?

Speaker 25:56

Yeah, um, I frankly have have been pretty surprised by the inertia of the per seat model. The per seat pricing, I think, has stuck around longer than than I expected. But uh yes, for our more kind of you know agentic vertical companies, it's often a per task type of pricing, right? So I mentioned some of our examples, right? So uh, you know, mortgage underwriting, Friday Harbor charges per loan, right? So per home loan that goes through their system, that's their pricing model. And that makes a lot of sense, right? That, you know, per loan is also how um these mortgage originators charge for their services. Um, you know, immigration, KCM is charging per visa filing, right? Very standard. That's how the customers are being charged. Um, and then CADI is quite interesting. So Caddy is an automation platform for professional services. So lawyers, right, financial advisors can actually screen record a workflow of something that they do repeatedly. And Caddy actually uses AI to understand the context of that screen recording, write deterministic code to actually connect all the APIs and backend tooling together and automate that workflow, which they call a loop. So they actually price per loop, which is quite interesting, right? Um so every workflow that they've automated, um, they charge for that. Um so I do think this per task um type of pricing makes a lot of intellectual sense, right? To customers, it aligns with the value and the revenue models that they have. The usage-based pricing, I think, had a moment, but I think that's more of a necessity, right? Because of the cost of the models and you know, costs, costs for um this like I guess not the the like latest models, right? Costs for prior gen models have definitely gone down. People are still paying a lot for the latest models. But I think pri pricing based on usage, like we haven't seen as a big thing in our incubator, is more task-based.

Speaker 2 27:58

Got it. So just to kind of double-click on this pricing thing. So the uh incumbents who have perpetuated the large incumbents who have perpetuated the perceiv model and who are also the at the moment the guardians of the large language models, and then therefore their own agents, et cetera. It's in their interest to continue to not cannibalize their current business model. What, you know, the perceived. They could extend it, they could, they could do a lot more, and it could be incremental. While as a startup, you have to somehow convince me as the buyer that that the outcome that I will get is dramatically better than if I did not do it, right? That's sort of the premise of this, either the task, or if you're a mortgage, it's task, and if you're uh immigration filing, you know, what whatever the case may be. Do you see that as the permeation of startups in these vertical applications break the foundation of the perceit model? That somewhere along they'll say, hey, you know, the transition has to be like when they went from pure licensing pay for a year and to then proceed, now they will go to tasks also? Or will they try to hang on? Because isn't their interest to hang on? There's no reason for them to give up.

Speaker 29:12

Yeah, I mean, that that's a really good question. You know, I I do think people are gonna try to hang on for as long as possible. I guess, you know, as a startup founder, we don't have the hangups of the legacy right system. But, you know, a lot of our founders are thinking about um, if we do our jobs right, maybe the number of seats in this organization might not grow as quickly as before or maybe even shrink, right? So if we do a per seat model, that revenue is actually going to shrink over time versus a per task model, which, you know, AI can hopefully do even more tasks. Um so from the a purely economical decision making, right? Like you kind of want to be on the task side versus the per seat side. I don't know how that will pressure the incumbents. You know, you think that it would pressure them to also align with task base pricing when the customers are also charging by the task versus, you know, by the receipt, right? By the number of employees, but it's hard to tell because I I do think there's a ton of inertia, right? Like people are used to what they're used to. This will probably stay the same for longer than we expect.

Speaker 2 30:20

Got it. Okay. And I also think that billing would become a little easier, right? Because either you completed the mortgage or you completed the filing, or the the task itself is is binary. Either it was completed or it wasn't completed, or there's not, I don't know, we had five employees, we have now 50 employees, we're down to 45 employees. You know, this this seems like it could be a very binary, clear thinking on hey, if I get it, I'll pay for it, if not. That seems like a easier from a for explanation to buyers. The risk is shared between the your startup and and the and the buyer.

Speaker 30:56

Yes, absolutely. And it puts pressure on the provider, right? In this case, a startup, to be able to actually deliver on that task, which at the end of the day is what everyone wants, right? We don't want to pay per se or like have another software. What we want is for a task to be done well.

Speaker 2 31:12

Yeah, yeah. So now the startups themselves are emerging all over the world, right? It used to be we had a corner on it, the Pacific Northwest, Silicon Valley had a corner. Either we had access to technology or access to talent, or usually both. Um, and also uh, you know, the ecosystem also was well developed for startups to emerge. It seems like now that's getting democratized, meaning every area of the world sees AI as the frontier, irrespective of whether or not they had a startup ecosystem, right? So is your incubator model transportable to other geographies? Or I know that you you guys are committed to Pacific Northwest. Is the do you see collaborations happening with governments or agencies that require the same level of uh uh you know and uh uh or some way for startups to emerge in their own geographies? Because everybody is not going to pick up and move from, I don't know, Middle East or Southeast Asia or uh or the Far East to come and start here, and there's all these other dynamics, right? So how do you see your expansion uh aspirations or collaboration aspirations?

Speaker 32:30

Yeah, um, so with AI2 incubator, actually during the pandemic, we started doing remote incubations. Um, and it actually ended up you know going pretty well. Now about 40% of our portfolio is typically remote incubations. So that could be, you know, founders are in SF or New York or Montreal, or maybe they're remote teams. Uh, they're coming into Seattle at least, you know, a week a quarter, you know, often more, but still, right, most of the incubation is remote. So a lot of our you know, programming, our round tables, right? Our experts are available for remote. Like we're all used to these Zoom calls now. So it's gotten a lot easier. Um, that being said, 60% of the portfolio is Seattle-based. And this year, I actually launched AI House, our physical space, right, for um this exact purpose um here in Seattle, because I very much believe in, you know, the value of Seattle's especially technical AI engineering talent. Today, I think a lot of that talent is up in the bigger companies here uh that you know create a huge cast over the startup ecosystem in Seattle. Um, but I do see that changing. And I think this is the moment, right? Seattle has uh about a quarter of all the AI engineers in the country. Uh SEP has 35% or at like 23%, I think. Um in every other city is well below, right? 10% or lower. And like I mentioned, in this age of a genetic AI, especially, but AI in general, right? And and AI, like I should say, we're not just investing in like LLM agentic AI. Like we have computer vision, we have robotics, we have, you know, all types, like custom models, right? Um, in healthcare, for example. We just think that building a great product today requires a level of engineering expertise that maybe wasn't required pre-this AI wave, right? And I I've been through the two ways, right? Mobile era. I remember the early days when building, you know, iOS apps was extremely difficult and only a certain number of people in the world knew how to do it well. Versus, right, my second company, Loftiam, I think, was in post-that era, right? When everything was kind of commoditized, like building web app, like app, like everything's like pretty easy. React native, like it systems were figured out, right? So, you know, the value of that engineering talent, I think, diminished a bit when um it was all flattened. We are not there in the AI era. We are so early. Um, and you know, the models are changing, right? Everything's getting more complex with agentec systems. It's actually a great time to be a city that has as much AI engineering talent as Seattle. And so that that's why, you know, I went all in on AI, Seattle AI, AI house. Um, I think now is the moment for us to really harness that, hopefully for startups, right? So the next generation, if you think AI startups, you know, hopefully a good number of those are are going to be coming from Seattle.

Speaker 2 35:21

Yeah, that that's an incredible statistic you just shared. 23% of all AI engineers live in the Pacific North. That's incredible to think that, you know, that's such a war for talent at the moment. And uh as uh as you know the big tech companies foster that talent, some of them will move on to start companies on their own. That it could be the Seattle's moment to, you know, change the narrative that it's harder to start companies here. Because that's been one of the knocks on Seattle is it's harder to start companies here because the big tech companies make it very comfortable for founders to come out. But looks like that is probably changing because the risk is a slightly different than ever before.

Speaker 36:05

Yes, definitely. And I I do think in Seattle it's nice to be running an you know AI2 incubator here in Seattle versus other markets like SF. You know, in SF and maybe to a certain extent New York, there's a bunch of startups, right? You can have your pick of going from like a Microsoft to uh maybe like a Series B startup to a Seed Stage startup, and then just starting your own company. So you can kind of size down and get more of that experience. In Seattle, it's hard, right? Like we don't have that whole cohort of startups. So we do see a lot of people going from big company, like um actually Priyanka from Kaesium went from Microsoft. She was in the Azure OpenAI team there at Microsoft, um, a researcher, um working with scientists. She went from that to being a founder directly. So that's a really big transition in everything related to your day-to-day. And like the skills required are very different. Um, I think that is where an incubator could provide a ton of value, right? Um, we helped her learn a lot of these skills from like recruiting these different types of people, like recruiting lawyers, negotiating deals, right? Fundraising, like product market fit. So yeah, it's it's nice, right? The incubator role, like a lot of people might think, like, oh, you know, why an incubator? Why not just be a direct investor? I do think it's very difficult to invest directly unless you have this group of people that not only have technical talent, but also have startup know-how. And we have the talent. I think the level of startup know-how in Seattle is unfortunately still very low. And that is, you know, my goal with the incubator and with AI House is that we can really level people up on startup knowledge. And then suddenly, hopefully, this ecosystem really explodes.

Speaker 2 37:53

Excellent. That I think is so true. We need more of these the casing around a startup because there are a lot of good founders, but there's more to founding a company. And as you originally said, it's the trough of adversity that hits the founder more often, the highs that they experience. And you can I can only imagine when your $30 million, $50 million revenue run rate just evaporates, right? So it requires some special, you know, special out-of-body experience almost.

Speaker 38:27

Yep. Yep. And you know, that hopefully won't happen to most people. I I do think my story is on the extreme side. I I was uh CUM residence with Techstars before they left Seattle, and I told my story to the founders, and I think someone in the team, Carson, said he had been there for four years, and mine out of all the Wednesday night like dinners was the craziest story that you'd ever heard. So hopefully there's not another pandemic that you know stops travel globally. But uh yeah, like you are gonna be hit with some impossible seeming challenges, right? But also the highs are the highest highs like you'll ever have. So, you know, that's why people do it, I guess.

Speaker 2 39:09

Yeah. Resilience is first the first virtue that you could have. So it's awesome. So my last question, what's your prediction? Maybe next three, five years. Well where do you you are you are now at the germination of everything that could happen in the world. What's your prediction for the next five years that likely to happen based on what you see? What's your Christian ball say?

Speaker 39:34

Likely to happen. You know, I I always think that things probably I think the models will improve very quickly. I actually am very bullish on that. And I think model improvement kind of underlies everything in AI, right? Um, you can only do so much in in terms of the application if the underlying models are not smart enough to actually do your job. So my prediction is actually like models will improve dramatically. Um, I think we're gonna be seeing all types of new interfaces, which is very exciting in AI. So, you know, voice interface is something that's very interesting right now to think about like how do we interact with AI with our computers? I think all of that will look very different in five years to the point where we'll we'll look back and think about like we used to type physically into a computer. Like that's you know, so inefficient. And I can't believe people used to do that. So yeah, uh, you know, beyond that, who knows?

Speaker 2 40:28

Well, first of all, uh, thanks for all the work you're doing. Very good luck with your new fund. And uh, we are very excited that uh we are able to collaborate with you as a as a nonprofit that fosters entrepreneurship. The relationship goes way back, and we are so excited to be uh, you know, be in the city where you are doing some very phenomenal work to grow entrepreneurship in general. So thank you for that. Sharisha, back to you. Great. Thank you so much for watching.

Speaker 40:57

You want to clarify uh our AT incubator is a for-profit entity. I have all the respect to the research entities is a non-profit. We are for-profit uh incubators. Yeah, uh, yeah.

Speaker 2 41:14

No, no, I'm saying Thai Seattle is a not-for-profit. Thai Seattle is a not-for-profit, not your entity. Thai Seattle is a not-for-profit because we we we are here to make sure we want to collaborate with the pro for-profit entities so that entrepreneurs find better ways to make uh make companies work. That's our goal.

Speaker 1 41:33

Amazing.

Speaker 2 41:34

Yeah. All right, all right.

Speaker 1 41:36

Thank you very much.

Speaker 41:38

Thank you and wish you all the best with your new fund. Thank you so much, both of you. This is really a lot of fun.

Speaker 2 41:44

Thank you for listening to our podcast from Startup Exit, brought to you by TiE Seattle. Assisting in production today are Eesha Jain and Minee Verma. Please subscribe to our podcast and rate our podcast wherever you listen to them. Hope you enjoyed it.