This video is a conversation recorded on August 23, 2025, featuring young venture investor Zoon Chang as a guest, delving deep into AI-era investment trends, opportunities in the Korean market, and the importance of community. In particular, it closely analyzes the success secrets of AI-native companies like Midjourney and Lovable that achieved explosive growth with small teams, and how top U.S. VCs are adapting their investment strategies for the AI era. It emphasizes that Korea has a unique opportunity to leapfrog directly to AI-native structures without legacy B2B SaaS systems, and how important it is to improve AI literacy across diverse job functions and build communities to achieve this.
1. How Today's Top-Performing VCs Are Moving
Today, top U.S. venture capital firms are moving in completely different ways in the AI space compared to the past. Similar to the crypto markets of 2018 and 2021-2022, everyone wanted to invest but there were too few 'makers' who understood the technology and could build — meaning capital provided very little differentiated value. Because legacy VCs lost their edge in those periods, all VCs are now going all-in on the AI era.
These VC movements can be broadly categorized into three types:
1-1. King Maker: Creating Great Founders
These firms focus on helping people become great founders. A prime example is AI Grant, which has achieved overwhelming success — reportedly most well-known AI-native services and products come from AI Grant alumni. They invest approximately $350,000 unconditionally and run batch programs connecting participants with leading AI figures and C-level executives from companies like Databricks, Snowflake, and Scale AI. Others like Cursor, Wispr, Neo, and South Common Park started as builder-centric communities and grew into influential investors riding the AI boom. They access talent and companies first, in a much denser way than Y Combinator used to.
"I think 'I want to get in there' is more accurate. Y Combinator is now accepting too many companies."
1-2. Game Maker: Building Exclusive Relationships
This type provides AI-specialized value to build exclusive relationships. Examples include creating and operating a fund jointly with Anthropic, or Sarah Guo and Elad Gil's podcast "No Priors," which invests in successful companies through relationships with top AI figures. VCs created by founders who built labs with Geoffrey Hinton, and UK-based venture firms that differentiate by providing academic rewards through research networks, also fall into this category.
1-3. Game Changer: Transforming Business Models Through AI Roll-ups
The most interesting aspect is that U.S. VCs are going beyond partnership structures to directly operate companies like holding companies while executing AI roll-up strategies. They recruit executives who held key roles in startups to pursue AI businesses together. For example, they acquire organizations like homeowners' associations and use AI to handle previously unautomated tasks, covering broader areas with fewer people. This raises operating margins and creates attractive enterprise value, essentially executing the roll-up business that PE (private equity) used to do, but in a new way using AI. The core idea is transforming previously uncompetitive or unattractive industries into highly attractive ones through AI. A representative example is rapidly acquiring call centers or customer service (CS) companies to increase operational efficiency and more than double enterprise value.
"Roll-ups were originally something PE firms did a lot. Instead of depending on individual founders' capabilities, you acquire, refactor, and satisfy the business needs for operating income, then sell to a buyer who can execute it."
1-4. BPO (Business Process Outsourcing): Handling Non-Core Tasks End-to-End
One of the most fascinating phenomena right now is BPO (Business Process Outsourcing). This involves performing other people's work on their behalf — data labeling, image-related OCR tasks, e-commerce product page creation, etc. While AI literacy has dramatically lowered knowledge labor costs, customers' willingness to pay remains high, creating huge opportunities in these areas. Satisfying early customers well means that even as competition intensifies later, economies of scale can help dominate mega-sectors. This connects to the AI roll-ups mentioned earlier — AI intervenes in traditionally labor-intensive value chains to improve efficiency, and the cost savings convert into margins. In the future, companies will retain only core competencies and outsource everything else, and when combined with the unique characteristics of the Korean market, this model has the potential to spread globally even faster.
2. The AI Era: Companies Destroying Corporate Structures and Their Astonishing Numbers
Today, AI technology is having a tremendous impact on the very concept of a corporation. Since the Industrial Revolution, the biggest bottleneck for companies has been knowledge labor. People skilled in knowledge labor created high added value, but now this is likely to change fundamentally. Companies that boldly break existing assumptions and structures are gaining the greatest growth opportunities. The proof is that early-stage startups are emerging with per-employee revenue levels comparable to Big Tech companies.
Looking at per-employee revenue of top AI startups is astounding:
- Cursor: Approximately 4.5 billion KRW per person
- Midjourney: Approximately 3 billion KRW per person
- Google: Approximately 2.6 billion KRW per person
- OpenAI: Approximately 2 billion KRW per person
The fact that companies exist that surpass Google is shocking. Whether this represents healthy growth remains to be seen, but the very possibility demonstrates the AI era's disruptive power. Particularly noteworthy is that there are 7 unicorn startups with fewer than 100 employees in the AI space — a key indicator of unprecedented change.
2-1. Midjourney: Building a Business on Discord
Midjourney designed its infrastructure structure and virtuous cycle brilliantly for explosive growth. Using Discord — a collaboration and community communication tool akin to Slack for gaming — they built their business foundation in a scalable form. They eliminated even login functionality, offloading communication infrastructure, real-time chat, and image upload loads entirely to Discord. This may seem like entrusting business fundamentals to an external party, but it actually meant choosing leaner, more efficient infrastructure and reaping the growth benefits.
Furthermore, Midjourney's core AI model was strengthened through Discord. Just as ChatGPT learns from user feedback like "do you prefer the left or the right?", Midjourney learned from which images users selected through Discord and continuously reinforced its model. This was Midjourney's fundamental competitive advantage.
They also employed a thoroughly community-based spread strategy. Free users on Discord could use the service for free, but generated images weren't hidden — they were visible to everyone. This allowed new users to see how others created images and follow along, generating powerful network effects.
"I thought that was a really smart strategy." "We got our feel for prompting in the Midjourney Discord in early '22. I learned a lot there too." "There was tremendous innovation within those constraints."
2-2. Lovable: Achieving 100 Billion KRW in Revenue with a Generalist Team
Lovable has attracted intense attention recently, recording over 100 billion KRW in revenue in just 10 months as a non-U.S. company. It's a visual coding tool combining the roles of full-stack software engineers and designers. At the time they surpassed 100 billion KRW in revenue, they had only 18 employees — 10 engineers, 3 in growth, and zero sales staff.
These remarkable results were achieved through:
- Conversational input-based customer communication: Starting as a development assistant tool, they created a virtuous cycle of continuously building the features customers truly wanted next through customer data and conversational input.
- Maximizing generalists: Traditional marketers would have emphasized ad creation or platform-specific expertise, but Lovable assembled generalists with the mindset of "I'll do all of this myself" — made possible because AI dramatically lowered the barrier to acquiring such know-how. The scope of work a single person could perform regardless of job title expanded enormously.
- Combining dogfooding and marketing: AI industry enthusiasts used their own products extensively, identified customer problems and bottlenecks, and naturally connected those insights to marketing. They rapidly grew the community through product-powered marketing — like generating a homepage instantly when you input a LinkedIn link.
"The scope of what one person can do regardless of job title has expanded enormously." "Exactly. And conversely, the need for internal communication decreases."
2-3. EvenUp: Handling Expert Knowledge Labor as Turnkey Service
EvenUp is a vertical AI company based on personal injury report fees. Similar to Korea's loss adjusters, it uses AI to directly resolve matters like determining compensation amounts after traffic accidents. Previously, everything had to be processed manually like accountants and tax consultants, but through AI-powered knowledge labor and modularization, they achieved tremendous efficiency and can provide services at massive turnkey scale. This enabled them to become a unicorn startup rapidly with around 100 employees.
2-4. Capabilities Needed for Today's AI-Native Organizations
Synthesizing these cases, Lean Startup methodology is gradually weakening. In the AI era, structural design and generalists who can perform diverse tasks well matter more than being lightweight, and a methodological obsession with how products reach customers is gaining new prominence. In the past, people thought "they succeeded because they built this product," but in the future, causality will shift to "this team will build a big startup because they understand this methodology well and can execute it tactically."
3. Three Ways Companies Approach AI
Companies' approaches to adopting AI can be broadly categorized into three generations:
- Companies that view AI as a tool and use it appropriately
- Companies that perceive AI as a crisis and try to respond and keep up
- Companies that see AI as an opportunity and aim to grow in fundamentally different ways from a structural perspective
The most important is obviously the third — seeing AI as an opportunity and rethinking from fundamentals. AI is an excellent means for expanding impact and creating something in the startup scene, so thinking from the most essential level is critical.
3-1. The Link Between Labor Costs and Performance Is Breaking
The traditional venture investment logic was: a small team or startup appears, proposes "we can grow well by doing this," investors agree and provide funding. However, the statistic that approximately 70% of early-stage startup funding goes to labor costs is no longer valid in the AI era. The model of receiving capital, hiring people, and creating a virtuous cycle of customer satisfaction is breaking down. When AI is used well, customer satisfaction can be improved with fewer people, changing the role of capital. This means the standardized success formulas and growth formulas in venture investing are collapsing.
3-2. Taste and Edge Become More Important Than Simple Knowledge Labor
The way VCs use AI in their work is also changing. Previously, AI was mainly used for relationship management or knowledge management across sourcing, due diligence, negotiation, investment, post-investment management, and exit processes. CRM has already become quite efficient, but knowledge management is undergoing tremendous change. VCs used to employ many interns to research industries and companies, but now deep research AI agents have largely replaced this role. Instead, interns who are good at Instagram — interns with taste and edge — have become much more valuable. Workers who build their own assets are now more important than those who simply provide knowledge labor.
"An intern with taste is more meaningful." "A person with edge." "That's right." "Someone who builds assets somewhere is far more valuable than someone who provides simple intellectual labor."
3-3. How to Organize AI When Pursuing Diversity
These changes lead to practical reflections on work methods. In industries like finance where the added value of judgment outweighs execution itself, if there's someone with an IQ of 200, a person with an IQ of 180 isn't very valuable. But if there's someone with an EQ of 150, they are valuable. Diversity in decision-making and various activities is what drives added value. An ideal investment firm should pursue such diversity, but it's not easy for all members to use AI well together. There's deepening deliberation about how to structurally leverage AI to gain virtuous cycles and advantages while absorbing all of this. This concern is shared not just by VCs but by all companies trying to use AI effectively.
4. Why VCs Are Creating AI Study Communities
VCs ultimately aim to discover great companies, invest at early stages, follow their growth stages, and earn returns. To do this overwhelmingly well, Zoon Chang referenced cases like AI Grant.
4-1. What Is a VC?
Zoon Chang defines the core of VC as creating value. Going beyond simply moving capital, he emphasizes that you should acquire equity with the mindset of becoming a startup's co-founder and contribute to building the company together. The VC industry had become too mature — formulaic and passive. Especially since the mid-2000s, as founder value was increasingly appreciated, VCs tended to provide capital and step back. But in this new era, he argues VCs must once again become entities that build things together.
4-2. What VCs Need to Attract Good Teams
For VCs to attract good teams and companies, various elements — good business opportunities, excellent teams, market conditions — must align. Traditional powerhouses like Kleiner Perkins and Sequoia were followed by Y Combinator creating a new wave, but now even Y Combinator is becoming old guard as new players keep emerging. This shows that the means of attracting good teams and founders are constantly evolving.
Especially as we enter the AI era, startup founders no longer blindly crave only capital. Various other factors have become important. So what should VCs provide to AI-native founders in the AI era?
"Ultimately, what do startups want most? I think it comes down to one thing: can you help us grow? So the feeling is 'if I get in there, there will be many growth catalysts for me.'"
4-3. Density of Problem-Solving Teams Is Needed
Startup CEOs and teams ultimately must solve their own challenges. The more they can hear from capable people about how others solved similar challenges — the trial and error, decision rationale, context, and results — the steeper their learning curve becomes. This is also why San Francisco holds an overwhelming advantage. Clustering begets more clustering, enabling very rapid and abundant innovation internally. In other words, density is critically important for gaining tacit know-how and real networks and help.
4-4. Talent with AI Literacy Is Needed
Another crucial element in the AI era is AI literacy. The biggest bottleneck in organizationally leveraging AI well right now is that people in non-developer roles are struggling alone at their positions, trying to figure out how to use AI. Existing AI communities and learning resources are too polarized. One side is too deep-tech for anyone who isn't a software engineer to access, and the other focuses on solopreneurs building businesses — there's a lack of knowledge exchange and verification platforms for people who want to "leverage this capability together within a good, impactful company."
4-5. A Single Domain or Job Function Isn't Enough — Diversity Is the Answer
When you ask "Are we really working 10x faster thanks to AI?", there's plenty of talk about how software engineers become more efficient, but far too little about all other job functions. Ultimately, you need to create business value as a team using this tool. Zoon Chang gathered people who share these concerns to form a community — to accelerate growth by sharing field successes and failures with each other. People from diverse job functions and positions applied, with a particular focus on non-developer roles.
"Ultimately, this needs to be leveraged well as a company — one person using it well alone has limited added value. For that, I need to deeply understand how other people in my company are using it, gather that knowledge... We're trying to compose the group as diversely as possible."
As someone who worked in talent development for many years noted, the area where talent development and capability enhancement are most numerically visible is healthcare. When a new treatment emerges, its impact on patients shows in the numbers. Over 90% of cases where experts break existing frameworks and 'unlearn' something new happen through interactions with people from other departments. Rather than worrying only within the same department, contemplation from similar but different perspectives is what ultimately advances society. This connects closely to the larger concept of AI, making us feel that diversity is an incredibly important value.
5. Why Building a Community "Right Now, in Korea"
Zoon Chang's purpose in building this community includes gaining current market intuition, but he's also keeping in mind several hypotheses: that the timing is right, and that Korea's strengths align to offer possibilities different from Silicon Valley.
5-1. Lack of Density for Growth! A Sense of Crisis
This movement actually started from a sense of crisis, he says. "I'm in Korea right now, but if I spent just one month in San Francisco I wouldn't be like this — that's how much I'd fall behind" — a thought that everyone working hard in their own position likely shares. The key point is that the density of stimulation is insufficient, and if it can be artificially increased, it absolutely should be. The ultimate goal is to keep pace with or maintain the lead.
5-2. First-Mover Advantages in Talent and Business Remain in the Korean Market
Conversely, the U.S. has already attracted so much talent that it's formed a perfectly competitive market where surplus profits are hard to generate. But in Korea, possessing these differentiated capabilities is much more likely to generate business surplus profits, which is why he personally feels it's more valuable to do these activities in Korea. While there's a perception that starting a business in Korea is difficult, paradoxically, specializing in a particular domain means competition is likely to be very low. Like building brands such as K-Beauty or K-pop idols, you can seize B2B opportunities vertically, one by one, entering in a very monopolistic fashion.
5-3. The Possibility of Leapfrogging from Bare Ground Directly to AI-Native
In the past, Korea was reluctant about B2B SaaS or transformative productivity improvements. This was because labor was too cheap and job security was high, so there was little felt need for unlearning and little willingness to change. This is also why many Korean companies lack well-documented manuals.
But Zoon Chang compares this to a society that used only cash leapfrogging past credit cards directly to QR payments. Just as the absence of the credit card intermediary actually made the transition easier and caused value to explode, Korea's lack of well-structured B2B SaaS could actually be a legacy-free starting point in the AI era. The U.S. has everything so well-structured — productivity and all — that 'unlearning' is difficult in some ways. The hypothesis is that Korea can have many companies that start directly with AI-native environments without such legacy.
"So the U.S. appears more advanced with lots of B2B SaaS and such, but from an AI-era perspective, that's actually acting as debt. Korea doesn't have that, so it can move much more easily toward AI-native processes." "Exactly."
The various bottlenecks that Korean society has — heading toward a low-birth-rate, aging society while still struggling with employment — could be greatly alleviated through AI technology. In the past, Korea easily utilized human resources due to an oversupply of knowledge labor, processing work without efforts to systematize. But now, in an era of increasingly scarce talent, many new businesses will emerge using AI technology. This emphasizes that it could be a new generation free from the legacy systems that countries like the U.S. built to use talent efficiently. Weaknesses are always connected to strengths, and this change aligns most closely with AI technology, offering the prospect that many Korean legacy companies will transform greatly through this opportunity and even expand globally.
5-4. What We Need Now: Individual Talent Ready to Unlearn & Learn
For these changes to happen, the single most lacking thing right now is people who are already using AI but want to use it better, gathering together to improve their capabilities and knowledge. Once that happens, the next opportunities and paths will naturally continue to open. This community also serves Zoon Chang's own learning purposes, while also identifying what problem awareness such talent holds.
Conclusion
Through today's conversation with young venture investor Zoon Chang, we gained deep insights into how VCs and companies are changing in the AI era, and particularly the unique opportunities that the Korean market holds. The remarkable results demonstrated by AI-native companies like Midjourney and Lovable demand a fundamental rethinking of corporate structure and operations, which is bringing major changes to traditional VC investment methods. VCs in the AI era must go beyond simply supplying capital to create value alongside founders and provide high-density environments for growth. Korea's weakness of lacking B2B SaaS legacy could actually become an opportunity to leapfrog directly into AI-native environments, and achieving this requires improving AI literacy across diverse job functions and building communities that enable active exchange. Ultimately, talent that is ready to embrace change through Unlearning and Learning and adapt to the new era will gather, stimulate each other, and learn — and this process will drive innovation. We look forward to Zoon Chang's proactive community-building efforts creating a positive ripple effect on Korea's AI ecosystem.