"Analyze all opinions in The Wall Street Journal from the past 3 months and write an Nvidia investment memo. Decide whether it will double in 3 years and also make a 3-page slide deck. In both Korean and English. Call Kevin to make a dinner reservation at the Westin Hotel for tomorrow at 7 p.m."
Seeing is believing. Wen Sang, co-founder and chief operating officer (COO) of the artificial intelligence (AI) startup Genspark (MainFunc Inc.), ran his personalized agent during the interview. His agent answered smoothly in English, then called Kevin speaking Korean. Kevin turned out to be Korean.
Starting in 2023 as a developer of next-generation AI search engines, Genspark boldly shifted its strategy early last year to become an agent developer. Users wanted results that went beyond information retrieval, such as compiling reports based on search results or executing code.
Genspark rolled out a personal agent product called "Super Agent" in Apr. 2025 and, in just five months, reached $50 million in annualized revenue. In November that year, it announced a $275 million (about 380 billion won) fundraising and became the first Silicon Valley agent company to reach unicorn status (corporate value $1.25 billion).
Sang, the COO, flatly said, "The era of humans being tormented by busy work is over," and added, "A world is coming where everyone works like a boss." Recently, Genspark also launched an enterprise agent product called "All-in-one Workspace." The following is a Q&A with him at the Korea AI Summit Seoul & Expo.
─ To be blunt, can you use AI agents to write a doctoral dissertation in a month?
"Why would that be impossible? Of course, dissertation work involves validation, but the writing process has completely changed. For example, Genspark can assign a fleet of assistants to a Ph.D. candidate.
There is an assistant who reads 100 papers and condenses the gist (TL;DR) into 30 words, an assistant who runs quantitative analysis models, and an assistant who drafts the thesis to meet university formatting requirements. It even helps with rehearsing oral presentations through a podcast feature. AI can also conduct surveys by making thousands of calls.
Humans review the outputs and act as the boss, giving directions like "make this part more concise" and "dig deeper into that analysis.""
─ If AI writes the dissertation, can a doctorate holder still be called a doctor?
"Let me use an analogy. We used to hand-wash clothes. Do we lament that our laundry skills have declined because we use washing machines now? In the past, meticulously finding and organizing literature was a yardstick of a researcher's diligence. Now that kind of low-level labor is for AI. Knowledge itself is democratized and available to anyone. Going forward, Ph.D. candidates will be judged on critical thinking and originality."
─ What are some examples in business settings?
"The way Scott, a solo consultant in Dallas, Texas, works is interesting. He helps restaurants secure loans. In the past, he hired people to count foot traffic and paid analysts, spending thousands of dollars over several weeks.
Now he pays Genspark just $25 a month and finishes in a day or two. Genspark does the local research, financial analysis, and even builds bank-facing slides. He now meets 10 more clients in the same amount of time. Every knowledge worker will work commanding a fleet, like the CEO of JPMorgan."
─ Not just work methods—social structures themselves could change.
"We are at a very early stage of a real transformation. What we experience now is only a tiny cross-section of a massive wave. An AI-native economic model that was previously unimaginable will emerge.
Think of Uber. Before smartphones and the internet, it was unthinkable to summon someone else's car with a single button. But once the infrastructure existed, we came to take that convenience for granted. The same goes for AI. Crossing the boundaries of traditional corporations and solopreneurs (one-person businesses), new collaboration structures and organizational forms will appear that are hard to even fathom today."
─ How is this different from the Industrial Revolution?
"If past civilization centered on factories and large organizations, the civilization we face may be reorganized around electricity, power, and tokens. With this new driving force, humanity might build rockets that traverse the solar system, conquer all diseases with technologies like AlphaFold, and manipulate matter freely to manufacture anything.
Perhaps humans could split into two broad groups. First, scientists—those who relentlessly push the frontiers of science and technology and expand human potential. Second, artists—those who expand the limits of expression through creativity and sensibility unique to humans.
The most fundamental change is that we no longer have to force ourselves to do what we don't want to do. For the first time in human history, we gain true choice and freedom. Now we can pick what we really want to do. Freedom—freedom."
─ Big Tech corporations are investing heavily in AI. What is Genspark's edge as a startup?
"Two things set us apart. First, the Mixture of Agents architecture. We are independent. We use Anthropic for coding, Google for images, and open-source models for simple tasks to produce optimal results.
Second, we focus on outcomes, not intelligence. OpenAI's benchmarks aim for intelligence that beats a chess grandmaster, but we prioritize reliability, usability, and accuracy in business settings.
Instead of solving Olympiad math problems, we build agents that write good emails, prepare investment memos, and persuade customers. Genspark is the arms and legs that handle real-world work.
Because we deliver tangible value, people are happy to pay. We are not a research company—we are a product company that generates revenue."
─ How does Genspark reduce hallucinations, the chronic issue of large language models (LLMs)?
"When one language model produces an output, several models with different characteristics cross-validate it. The mixture-of-agents structure drastically reduces hallucinations. On top of that, our evaluation benchmark kicks in—a feedback algorithm that gives high scores for good outputs and low scores for poor ones. The know-how of founders who designed Microsoft's search engine and TikTok's recommendation algorithm is embedded in it."
─ Talk of an AI bubble is also intense.
"A bubble emerges when the gap between perception and reality is too wide. If it's all talk and there are no real results, that's a bubble. But we're different. A voice agent making calls to book reservations and producing investment analysis reports in real time isn't fantasy—it's happening now. What's more, we launch new products every week. This pace would be impossible without AI. More than 80% of the code we write is built with AI.
I see it less as an AI bubble and more as a bottleneck in the technology cycle. As Satya Nadella, Microsoft's chief executive officer (CEO), recently noted, the infrastructure (chips) is in place, but there isn't enough power and data center capacity to run it. Such mismatches occur in economic cycles. Each time this bottleneck is resolved, AI's reach into our lives will expand by 30%, then 50%."
─ What are your plans for doing business in Korea?
"Judging by enthusiasm for AI adoption alone, Seoul is far hotter than Silicon Valley. From the practitioners and customers I met at the conference to the level of questions you're asking now, I can feel how deeply Korea understands this technology.
LG is a major investor in Genspark, and we are in talks with many large Korean corporations about partnerships. We will soon establish a base in Seoul and aggressively expand services optimized for the Korean market."